This retrospective cohort study included consecutive patients treated at a tertiary referral centre between January 2017 and January 2025. Eligible patients had late FRI after tibial or femoral intramedullary nailing and were treated using a staged reconstruction protocol based on multi-regional debridement and fixation replacement.

FRI was diagnosed using the international consensus framework, integrating clinical, microbiological, and histopathological evidence (Metsemakers et al., 2018). Clinical confirmatory findings included either a sinus tract or wound breakdown communicating with bone or implant, or purulent drainage or intraoperative pus. Microbiological confirmation required phenotypically indistinguishable microorganisms isolated from at least two separate deep specimens. Histopathological reports were reviewed in all patients. In late-onset cases, histopathological confirmation was defined as the presence of more than five polymorphonuclear neutrophils per high-power field in at least five separate high-power fields (Govaert et al., 2020). For time-to-onset classification, the interval from fracture fixation to the first documented onset of infection-related symptoms or signs was categorized according to the Willenegger–Roth framework as early onset (< 2 weeks), delayed onset (3–10 weeks), or late onset (> 10 weeks) (Metsemakers et al., 2020; Willenegger and Roth, 1986). Separately, for eligibility and descriptive analyses, the documented FRI interval was defined as the time from the earliest documented confirmation of FRI to admission for Stage I treatment at our centre. Patients were included if they met the following criteria: (1) previous tibial or femoral fracture treated with intramedullary nailing; (2) FRI diagnosed according to the criteria described above, with a documented FRI interval of at least 3 months before admission; (3) fracture nonunion; (4) treatment with the staged protocol described below; and (5) age between 18 and 60 years. Exclusion criteria were Cierny–Mader physiological class C host status, severe immunodeficiency or uncontrolled autoimmune disease precluding staged reconstruction, and follow-up of less than 12 months.

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Medical Ethics Committee of the 920th Hospital of the Joint Logistics Support Force of the PLA. Informed consent was obtained from all participants.

The primary outcome was confirmed infection recurrence after definitive Stage II reconstruction. Confirmed recurrence required at least one confirmatory FRI criterion: a sinus tract or wound breakdown communicating with bone or implant, purulent drainage or intraoperative pus, phenotypically indistinguishable microorganisms isolated from at least two separate deep specimens, or fulfilment of the predefined histopathological criterion (Metsemakers et al., 2018). Clinical concern for recurrent infection without fulfilment of a confirmatory criterion was recorded separately as suspected recurrence. Any unplanned surgical debridement for suspected or confirmed infection after definitive Stage II reconstruction was classified as post-reconstruction treatment failure.

Before definitive Stage II reconstruction, confirmed persistent infection was defined using the same FRI confirmatory criteria. Intraoperative warning findings that prompted additional sampling, repeat debridement, and postponement of bone grafting but did not fulfil a confirmatory criterion were classified as suspected residual infection. Confirmed persistent infection and suspected residual infection before definitive reconstruction were recorded as interstage treatment failures and were not counted as post-reconstruction recurrence.

Secondary outcomes included repeat debridement before final infection control, radiographic bone union, time to full weight-bearing, complications, inflammatory markers, and functional outcomes. Bone union was defined as bridging callus across at least three of four cortices on anteroposterior and lateral radiographs, together with the absence of pain or abnormal motion at the reconstruction site (Corrales et al., 2008). Time to full weight-bearing was defined as the interval from definitive Stage II reconstruction to the first documented follow-up visit at which unrestricted full weight-bearing was permitted by the treating surgeon. Inflammatory markers included white blood cell count, erythrocyte sedimentation rate, and C-reactive protein. These markers were evaluated as non-specific longitudinal adjuncts rather than as stand-alone tests to confirm or exclude infection. Functional outcomes included the Visual Analogue Scale (VAS; range 0–10, with higher scores indicating greater pain), Lower Extremity Functional Scale (LEFS; range 0–80, with higher scores indicating better lower-limb function), 20-item Self-Rating Anxiety Scale (SAS) raw score (range 20–80, with higher scores indicating more severe anxiety symptoms), and Short Form-36 (SF-36) total score (calculated as the sum of the eight domain scores, range 0–800, with higher scores indicating better health-related quality of life).

All patients were treated using a two-stage protocol. In Stage I, the intramedullary nail was removed, and systematic multi-regional debridement was performed in all patients. The fracture/nonunion site, nail entry site, entire medullary canal, and proximal and distal locking screw tracts were routinely treated as components of the potential nail-related infection pathway (Fig. 1b). Necrotic soft tissue, fistula tracts, sequestra, inflammatory granulation tissue, and purulent material were removed. The medullary canal was sequentially reamed using the diameter of the removed intramedullary nail as a reference. The final reaming diameter was individualized according to medullary canal morphology and cortical integrity but was at least 2 mm larger than the removed nail diameter. All proximal and distal locking screw tracts were routinely inspected and debrided along their entire trajectories. A tract was enlarged when gross loosening of the corresponding locking screw, visible inflammatory granulation tissue, purulence, or other macroscopic evidence of infection was present. Enlargement was performed using a drill bit with a diameter larger than that of the removed locking screw. If drilling alone did not adequately remove the involved tissue, the tract was additionally curetted. Irrigation and negative-pressure suction were performed through the treated tract to remove residual debris and permit inspection of the tract and adjacent medullary canal. Debridement was continued until no visible inflammatory tissue, purulence, or loose debris remained along the tract. No dedicated continuous irrigation-aspiration system or pressurized lavage device was used. After reaming, sterile normal saline was manually injected into the medullary canal through a standard sterile suction tube using a syringe, without simultaneous negative-pressure suction. After each irrigation cycle, the suction tube was introduced through the treated distal locking screw tract and connected to negative-pressure suction to remove residual fluid and debris. The canal was then reinspected. Reaming, curettage, irrigation, and intermittent suction were repeated as required until no visible inflammatory granulation tissue or purulent material remained and the irrigation effluent was clear of macroscopic debris. Additional cortical windows or longitudinal grooves were created in 14 patients when the adequacy of intramedullary debridement could not be confirmed through the nail entry site and locking screw tracts alone.

After debridement, mechanical stability was restored using locking plate fixation. All femoral cases received internal plate fixation. In tibial cases, internal or externalized plate fixation was selected according to the local soft-tissue condition. Screw placement across areas of gross purulence, necrotic bone, or the planned grafting zone was avoided whenever possible. Polymethyl methacrylate cement was manually mixed with vancomycin at a fixed concentration of 5 g per 40 g of cement powder. The vancomycin-loaded cement was used for dead-space management, local Gram-positive antimicrobial delivery, defect filling, and induced membrane formation (Fig. 1c) (Metsemakers et al., 2019). The cement rods and defect-filling cement were non-load-bearing local antibiotic carriers and spacers rather than cement-coated intramedullary fixation devices; mechanical stability was provided by the locking plate construct. The intramedullary cement rod was manually moulded around either a single Kirschner wire (K-wire) or two K-wires twisted together. The diameter of the cement rod did not exceed that of the removed intramedullary nail. The K-wire core extended beyond the cement mantle at the insertion end to facilitate subsequent removal. A single continuous cement rod was introduced through the original nail entry site and advanced distally within the medullary canal. Additional small cement rods were placed in the enlarged locking screw tracts. For large-diameter proximal fixation tracts associated with cephalomedullary nails, the tract-specific cement rod was likewise moulded around a K-wire core. At Stage II, the intramedullary cement rod was withdrawn through the original nail entry site, whereas the tract-specific cement rods were removed through their respective tracts. The remaining cement was used to fill the bone defect and cover the fixation construct when appropriate. Soft-tissue defects were managed with flap coverage when primary closure was not feasible.

Definitive Stage II reconstruction was not performed at a fixed time point. Patients were scheduled for a planned Stage II procedure no earlier than 6 to 8 weeks after Stage I and only after clinical and laboratory infection quiescence. The preoperative criteria included healing of the incision or fistula tract; absence of local signs of infection; and normalization or near normalization of white blood cell count, erythrocyte sedimentation rate, and C-reactive protein on repeated assessments. No single serum inflammatory-marker cutoff was used in isolation; results were interpreted according to serial trends and institutional reference ranges together with wound healing and soft-tissue findings (Metsemakers et al., 2018, 2020).

At the planned Stage II procedure, the Stage I locking plate and all vancomycin-loaded cement rods and spacers were removed. Repeat debridement was performed, and the previous screw tracts were inspected, curetted, and irrigated. The operative field was reassessed before bone grafting. Purulence was regarded as confirmatory evidence of active infection (Metsemakers et al., 2018), whereas unexpected turbid or excessive fluid accumulation and abnormal inflammatory granulation tissue of questionable quality or viability were treated as intraoperative warning findings.

When these findings raised concern for residual infection, multiple deep-tissue specimens were obtained for microbiological culture, and intraoperative frozen-section histopathological examination was performed. The decision to defer bone grafting and perform further debridement was based on the macroscopic findings and was made irrespective of the frozen-section result. When the decision to defer reconstruction was made after removal of the Stage I plate, temporary plate fixation was re-established to maintain mechanical stability during the additional infection-control interval. Definitive reconstruction was rescheduled after clinical and laboratory infection quiescence and soft-tissue readiness were re-established (Metsemakers et al., 2020).

When definitive reconstruction proceeded, the existing temporary plate construct was removed, bone grafting was performed, and definitive stability was restored using a new internal locking plate. No intramedullary nail was reinserted (Fig. 1d). Autologous iliac cancellous bone was used for grafting. If the autologous graft volume was insufficient, morselized allograft was added as an extender, with the allograft proportion limited to no more than 25 % according to the institutional protocol (Pesciallo et al., 2021).

Whenever clinically feasible, systemic antibiotics were discontinued for 1 to 2 weeks before Stage I surgery to improve microbiological yield. In two patients with systemic sepsis or rapidly progressive infection, antibiotics were continued and no preoperative antibiotic-free interval was used. At Stage I, 4.34 ± 0.96 deep specimens were obtained per patient before empirical postoperative antimicrobial therapy was initiated; the median number was 5 (interquartile range 4–5; range 2–5). Sampling included deep soft tissue and at least one lesional bone specimen in all patients. Multi-site deep soft-tissue sampling was performed in 16 patients. Representative sampling sites included the fracture site, medullary canal, nail entry site, and locking screw tracts.

Each specimen was placed in a separate sterile container and promptly transported to the microbiology laboratory. The specimens were inoculated onto agar media and incubated in a microbiological incubator. Separate instruments were not routinely used for each specimen. Aerobic cultures were incubated for 5 d. In cases without aerobic growth, anaerobic cultures were maintained for 14 d. Molecular testing was not routinely performed. Histopathological examination was performed in all patients. Additional details are provided in Table S3 in the Supplement.

Pending culture results, empirical intravenous treatment consisted of ceftriaxone 2 g once daily. Treatment was subsequently adjusted according to the identified organism and susceptibility profile. All patients received a total of 42 d of systemic antimicrobial therapy after Stage I, consisting of 14 d of initial intravenous treatment followed by 28 d of susceptibility-directed continuation therapy. Continuation therapy was administered orally when an active oral agent was available and intravenously when no active oral option was available. In culture-positive cases, antimicrobial agents were selected according to the identified organism and susceptibility profile. In culture-negative cases, the institutional regimen consisted of intravenous ceftriaxone 2 g once daily for 14 d, followed by oral levofloxacin 500 mg once daily plus rifampicin 450 mg once daily for 28 d. This regimen reflected institutional practice during the study period. The antimicrobial strategies are summarized in Table S1 in the Supplement.

Whenever repeat debridement was required before definitive Stage II reconstruction, multiple deep-tissue specimens were obtained before antimicrobial treatment was restarted. After repeat debridement, both affected patients began a new 6-week systemic antimicrobial course consisting of 14 d of intravenous treatment followed by 28 d of oral treatment. Patients who proceeded to definitive Stage II reconstruction without evidence of active infection received routine surgical prophylaxis with a single pre-incision dose of cefazolin; no postoperative therapeutic antimicrobial course was administered.

Vancomycin-loaded cement was regarded as an adjunct to debridement, dead-space management, local antimicrobial delivery, and systemic antimicrobial therapy, not as a stand-alone antimicrobial treatment (Metsemakers et al., 2019; Pesciallo et al., 2021; Depypere et al., 2020a).

Patients were followed monthly during the first 3 months after definitive Stage II reconstruction, every 3 months during the first postoperative year, every 6 months during the second year, and annually thereafter. At each follow-up visit, clinical signs of infection recurrence were assessed, including redness, swelling, increasing pain, wound drainage, or new sinus tract formation. White blood cell count, erythrocyte sedimentation rate, and C-reactive protein were reviewed. Radiographs were obtained to evaluate bone union. Functional scores were assessed at initial admission and final follow-up.

Statistical analyses were performed using IBM SPSS Statistics version 25.0. Continuous variables were assessed for normality using the Shapiro–Wilk test. Normally distributed variables are presented as mean and standard deviation, and non-normally distributed variables as median and interquartile range. Categorical variables are presented as numbers; percentages are provided only when clinically informative. Paired comparisons between initial admission and final follow-up were performed using the paired-samples t-test or Wilcoxon signed-rank test, as appropriate. A two-sided p value < 0.05 was considered statistically significant.

A total of 47 consecutive patients with late FRI after intramedullary nailing were included. There were 39 male and eight female patients, with a mean age of 40.2 ± 11.4 years. The mean documented FRI interval from the earliest documented confirmation of FRI to admission was 15.3 ± 15.7 months. All patients had a documented FRI interval of at least 3 months and fracture nonunion. Based on the interval from fracture fixation to the first documented onset of infection-related symptoms or signs, all infections were classified as late-onset FRI. The infected bone was the tibia in 21 patients and the femur in 26 patients. Eight patients had open fractures and 39 had closed fractures. Recorded comorbidities or risk factors included diabetes mellitus in nine patients, smoking in 17, immunosuppression or glucocorticoid use in three, and deep vein thrombosis in two (Table 1).

Histopathological examination was performed in all 47 patients, and all pathology reports described changes compatible with osteomyelitis. Of the 17 culture-negative patients, 15 fulfilled clinical confirmatory criteria based on a communicating sinus and/or intraoperative purulence. The remaining two patients fulfilled the histopathological confirmatory criterion, with more than five polymorphonuclear neutrophils per high-power field identified in at least five separate high-power fields. Deep-tissue cultures were positive in 30 patients and negative in 17. The distribution of identified organisms and multidrug resistance is summarized in Table 2.

After debridement, 21 patients had segmental bone defects and 26 had local bone defects. The mean segmental defect length was 6.7 cm, with a range of 4 to 13 cm. All patients were stabilized with locking plate constructs after Stage I debridement, including internal plate fixation in 37 patients and externalized plate fixation in 10. Additional cortical windows or longitudinal grooves were required in 14 patients. Flap coverage was required in four patients.

The mean interval between Stage I and definitive Stage II reconstruction was 11.4 weeks (range 7–22 weeks). At definitive Stage II reconstruction, the existing temporary plate construct was removed and replaced with a new internal locking plate in all 47 patients; no intramedullary nail was reinserted. Reconstruction was not performed at a fixed interval; timing was determined by wound healing, the absence of local signs of infection, inflammatory-marker trends, soft-tissue readiness, and the need for repeat debridement.

Two patients required repeat debridement before definitive Stage II reconstruction. One patient developed fever, wound drainage, and elevated serum inflammatory markers during the interstage period. At repeat debridement, five deep specimens were obtained, four of which yielded phenotypically indistinguishable Staphylococcus aureus. Histopathological examination demonstrated dense neutrophilic infiltration. This event fulfilled microbiological confirmatory criteria and was classified as confirmed persistent infection before definitive Stage II reconstruction.

The other patient had no preoperative clinical or laboratory evidence of active infection. At the planned Stage II procedure, removal of the Stage I plate and cement spacer revealed unexpected turbid or excessive fluid accumulation and abnormal inflammatory granulation tissue in the absence of purulence. On the basis of these macroscopic findings, bone grafting was deferred immediately, repeat debridement was performed, and temporary plate fixation was re-established to maintain mechanical stability during the additional infection-control interval. Multiple deep-tissue specimens were obtained for microbiological culture, and an intraoperative frozen-section histopathological examination was performed, with the result available during surgery. Five deep specimens remained culture negative; however, final histopathological examination demonstrated more than five polymorphonuclear neutrophils per high-power field in at least five separate high-power fields. The event therefore fulfilled the histopathological confirmatory criterion and was classified as confirmed persistent infection.

Both patients subsequently completed a new 42 d systemic antimicrobial course and met the predefined infection-quiescence criteria. They then underwent definitive Stage II reconstruction. In the second patient, the interim plate was removed and replaced with a new internal locking plate at the rescheduled procedure.

A representative case showing preoperative sinus formation, infected nonunion, Stage I multi-regional debridement, Stage II reconstruction, radiographic union, and final functional recovery are shown in Fig. 2.

The mean follow-up after definitive Stage II reconstruction was 25.3 ± 14.1 months, with a range of 12 to 52 months. No confirmed infection recurrence or unplanned surgical debridement for suspected recurrence occurred after definitive Stage II reconstruction. Outcomes stratified by microbiological category are shown in Table S2 in the Supplement.

Union after the index Stage II reconstruction was achieved in 44 of 47 patients (93.6 %). Three patients, including two femoral cases and one tibial case, had persistent nonunion and underwent repeat bone grafting with or without fixation revision. All three subsequently achieved union, resulting in a final union rate of 100 %. The mean time to radiographic union was 9.7 ± 2.1 months, with a range of 5 to 11 months. Full weight-bearing was achieved at a mean of 11.3 ± 3.8 months, with a range of 8 to 17 months. Delayed iliac donor-site wound healing occurred in five patients; all events were superficial and resolved with local wound care without surgical re-intervention. The specific causes of delayed healing could not be reliably determined from the available retrospective medical records. Postoperative knee stiffness occurred in five patients with femoral infection. Limb length discrepancy occurred in three patients, measuring 2.7, 3.0, and 4.3 cm.

WBC count, ESR, and CRP decreased; LEFS and SF-36 scores increased; and VAS and SAS scores decreased between initial admission and final follow-up (WBC, p=0.001; all other comparisons, p<0.001; Table 3).