All adult patients with mandibular OCF-related osteomyelitis followed up in our reference center for the management of complex BJI between 1 September 2012 and 31 July 2019; they were included in a retrospective observational cohort study. Infections requiring less than 1 month of antimicrobial therapy were considered skin and soft tissue infections (SSTIs), so these cases were excluded.

For each patient, data were collected from medical records and biological software in an anonymous standardized case report form. Co-morbidities were summarized by the Charlson co-morbidity index and the American Society of Anesthesiologists (ASA) score.

The harvesting step was performed through an anterior or posterior approach, with or without pneumatic tourniquet, depending on the operator's habits, using 2.0 magnifying glasses. A dose of platelet aggregation inhibitor was administered to the patient just before flap ischemia. Perforators were located visually in the distal third of the leg at the beginning of the procedure, with non-routine help of an ultrasound Doppler. The skin paddle was positioned with the most suitable perforator (size, pulsatility) in its center. If possible, two perforators were included. The skin paddle could be set intraorally or extraorally depending on the defect to be reconstructed, without tension on the suture line. If both sides had to be reconstructed, we used the skin paddle for extraoral reconstruction and a muscle for the intraoral side (either the tibialis posterior or a part of the soleus). The donor vessels were the fibular artery which divided just below the bifurcation of the tibiofibular trunk and one of the fibular veins. Arterial anastomoses were performed in a termino-terminal fashion using separate stitches of 8-0 or 9-0 non-absorbable sutures (Ethilon™, Ethicon Inc., Scotland, 1915). Venous anastomoses are either performed in a termino-terminal fashion on one of the proximal branches of the thyro-linguo-facial trunk or in a termino-lateral fashion on the internal jugular vein in second intention, with separate stitches of non-absorbable thread. Regarding bone reconstruction, most of the patients included in the study benefited from preoperative digital planning, custom-made cutting guide, and patient-specific plates with load sharing osteosynthesis. After surgery, flaps were monitored clinically to detect early microvascular failure, using the skin palette as control (heat, color, skin recolouration time). Additionally, Doppler flowmetry monitoring was performed twice daily during the postoperative period, during the entire stay (20 d on average).

In the absence of consensual definition of OCF-associated osteomyelitis, diagnosis was based upon clinical, biological, imaging evidence, and/or culture of microorganisms from deep surgical samples, according to the definition of surgical site infection of the US Centers for Disease Control and Prevention (CDC) (Mangram et al., 1999).

On the basis of prosthetic joint infection chronological definitions (Kapadia et al., 2016; ICM, 2022), the delay between inoculation and symptoms was defined as early (≤3 months), delayed (3–12 months), and late (>12 months) postoperative infections. Similarly, infection was classified as “acute” when the delay between inoculation and management was ≤4 weeks.

All intraoperative tissue samples were inoculated onto a Columbia sheep's blood agar plate (read at days 1 and 2), two PolyVitex chocolate agar plates (read at days 1 and 2 for one plate and days 7 and 10 for the other), two blood agar plates for anaerobic incubation (read at days 3 and 5 for one plate and at days 7 and 10 for the other), and into a Schaedler anaerobic liquid broth. The broth was systematically subcultured on day 10 onto chocolate and blood agar plates for anaerobic incubation for 5 d. Identification and antibiotic susceptibility testing were performed following standard laboratory procedures (VITEK 2 system or VITEK MALDI-TOF MS; disk diffusion methods or ATB ANA device; bioMérieux, Marcy l'Etoile, France). Microbiological documentation was described, considering all virulent pathogens (i.e., Staphylococcus aureus, streptococci, and Gram-negative bacilli (GNB)) yielded out of deep surgical samples, and low virulent pathogens and/or potential contaminants (i.e., coagulase-negative staphylococci (CoNS), Corynebacterium spp., and Cutibacterium spp.) if present on at least two deep surgical samples, which is considered the gold standard (Osmon et al., 2013).

Antimicrobial therapy was considered empirical in the absence of previous microbiological documentation. It was considered appropriate when all pathogens highlighted by gold-standard sample cultures were targeted by at least one molecule.

Treatment failure was defined as symptom persistence under treatment (i.e., worsening or recurrence of local inflammatory symptoms, nonunion, sinus tract, abscess, purulent discharge, and bone and/or device exposure), infection relapse after treatment disruption, infection-related requirement of additional surgical procedure, flap loss, and infection-related death.

Descriptive statistics were used to estimate the frequencies of the study variables, described as numbers (%) for dichotomous values and medians (interquartile range, IQR) for continuous values. For the percentage calculation of each variable, the number of missing values was excluded from the denominator. Nonparametric statistical methods were used to compare the study groups (chi-squared test, Fisher's exact test, or Mann–Whitney U test, as appropriate). Determinants of treatment failure were assessed using: (i) logistic regression analysis, expressed as odds ratios (OR) and their 95 % confidence intervals (95 % CI), and (ii) Kaplan–Meier curve analysis representing the probability of treatment-failure-free survival with time, compared between groups using the log-rank (Mantel–Cox) test. A p value of ≤0.05 was considered significant. All analyses were performed using SPSS software version 17.0 (SPSS, Chicago, IL, USA) and GraphPad Prism version 5.3 (GraphPad Software, San Diego, CA, USA).

Forty-eight patients with mandibular OCF-associated osteomyelitis were included. Thirty of the 48 included patients (62.5 %) were male, with a median age of 60.5 (IQR, 52.4–66.6) years and a median modified Charlson co-morbidity index of 4 (IQR, 3–5).

Flap reconstruction was mostly performed for carcinoma (n=27/48; 56.3 %), and 34 out of 41 patients (82.9 %) had previous neck irradiation. The major donor flap was fibula (n=46/48; 95.8 %), with the two last patients having received clavi-pectoral and serrato-costal flaps. Cervical lymphadenectomy and tracheotomy were performed in 23/48 (47.9 %) and 24/48 (50.0 %) patients, respectively. Antimicrobial prophylaxis followed the French guidelines in 42 (97.7 %) of the 43 evaluable patients, mostly based on amoxicillin–clavulanate.

Patients and index surgery characteristics are summarized in Table 1.

OCF-related osteomyelitis was mostly early (n=43/48; 89.6 %) and was considered acute in 22/48 (45.8 %) cases. Main symptoms are presented in Table 1. Fever was present in 17/47 (36.2 %) patients. The median maximum CRP level was 90.3 mg L-1 (IQR, 34.2–158.5). Clinical symptoms were associated with radiological signs for infection in 75.0 % of cases (n=33/44).

Gold-standard surgical samples were collected in 41/48 (85.4 %) patients, with a median of three (IQR, 1–5) samples per patient, allowing for a microbiological documentation in 40/41 (97.6 %) cases. Only three documented infections were monomicrobial. Main implicated pathogens were Enterobacteriaceae (n=25/41; 61.0 %), streptococci (n=22/41; 53.7 %) with a majority of milleri group (n=14/41; 34.1 %), Staphylococcus aureus (n=10/41; 24.4 %), Enterococcus faecalis (n=8/41; 19.5 %), anaerobes (n=8/41; 19.5 %), and non-fermenting GNB (n=8/41; 19.5 %). Only one patient had positive blood cultures (Actinomyces odontolyticus). Note that results of preoperative superficial microbiological samples were available in 26/48 (54.2 %) patients. When comparing gold-standard and superficial sample culture results in the 20 (41.6 %) patients for whom both were available (Table 2), only four (20.0 %) were fully consistent. A 100 % predictive negative value was found for CoNS, only. Conversely, specificity and predictive positive values were good for Enterobacteriaceae and Enterococcus faecalis.

Histopathological analysis was performed in 24/48 cases (50.0 %) with histological signs of infection in 12/22 (54.5 %) cases including chronic inflammation in 8/22 (36.4 %).

Out of the eight patients with no gold-standard microbiological documentation, all had local clinical and radiological signs of infection; four patients had histologically proven osteomyelitis, and six patients had local bacteriological samples highlighting invasive pathogens (milleri group streptococci, Enterobacteriaceae, and/or Pseudomonas aeruginosa).

Thirty-nine (81.3 %) of the 48 included patients benefited from surgery, mostly consisting of debridement with implant retention (DAIR; n=25/39; 64.1 %). Flap had to be removed in 7/39 (17.9 %) cases during the initial septic surgery.

Empirical antimicrobial therapy was given for 16.5 (IQR, 7.8–31.8) days and considered appropriate in 33/48 cases (68.8 %). The most commonly empirical regimen (n=22/48; 45.8 %) was a combination of a broad-spectrum beta-lactam targeting non-fermenting GNB (piperacillin–tazobactam, cefepim, or carbapenem) with either vancomycin or daptomycin. Anaerobes were empirically targeted in 47/48 (97.9 %) patients. Four (8.3 %) out of 48 patients empirically received an antifungal drug. After empirical antimicrobial therapy, the antimicrobial therapy was adapted to the definite microbiological documentation, with a complete oral administration possible in 16/43 (37.2 %) patients, only. Total duration of antimicrobial therapy was 93 (IQR, 64.0–127.5) days. Antimicrobial therapy was driven by an infectious disease (ID) specialist in 44/48 (91.7 %) cases, with a mean delay from symptoms to referral of 2.6 (IQR, 0.1–13.2) weeks. ID specialist advice was taken before surgery in 18/39 (46.1 %) of the operated patients.

In a median follow-up of 18 (IQR, 11.2–31.0) months, 24/48 (50.0 %) treatment failures were observed, corresponding to symptom persistence (n=16/48; 33.3 %), relapse (n=5/48; 10.4 %), requirement for additional surgical procedure (n=20/48; 42.6 %) including 4/48 (8.3 %) flap removal, and 3/48 (6.3 %) infection-related death.

Baseline characteristics and infection diagnostic criteria were not different between the subsets of patients with or without treatment failure (Table 1). A higher proportion of patients treated with DAIR were observed in the treatment failure subset (n=16/24; 66.7 % versus 9/24; 37.5 %; p=0.043), which consequently appeared as a determinant of treatment failure (OR, 3.333; 95 % CI, 1.020–10.898, p=0.046). The only other determinant of treatment outcome was an early referral to ID specialist (OR, 0.236 if ≤2 weeks; 95 % CI, 0.062–0.933; p=0.035). Non-fermenting GNB infections tended to be associated with a higher risk of failure (OR, 8.400; 95 % CI, 0.927–76.151; p=0.058) (Table 1). Kaplan–Meier curve analysis confirmed these results and additionally suggested the non-negative state of the biological inflammatory syndrome after 2 weeks as another potential predictor of treatment failure (Fig. 1).

Note that the eight patients with no gold-standard microbiological diagnosis received 92 (IQR, 69.8–112.0) days of antimicrobials, after a surgery in four patients. Persistent infection was noted in three patients, including two requiring an additional surgery and one death.

Concerning the nine patients who did not benefit from surgery, all had early postoperative infection. They all received an appropriate antimicrobial therapy, based on the result of multiple local samples, available for all patients, and guided by an infectious disease specialist, for a total duration of 98 (IQR, 86–127) days. No treatment failure was observed after a follow-up of 15.3 (IQR, 6.2–26.0) months after the end of treatment. Two patients died of non-infectious causes. No specificity could be highlighted in this specific subset of patients.