The Geneva University Hospital system is the only public hospital in Geneva and the tertiary center for parts of neighboring France. The institution has been using a clinical pathway [1] for DFUI since 2013, which includes input and consultation from diabetologists, podiatrists, angiologists, infectious diseases physicians, orthopedic and foot surgeons [4], radiologists [8], vascular surgeons, and specialized physiotherapists and surgical nurses (AT) for ortheses and patients’ education.
Definitions and study criteria
The diagnosis criteria for diabetes mellitus for this study stem from the American Diabetes Association and basically included a glycolysed hemoglobulin A1 level of ≥6.5%, a fasting glycemia level of 7.0 mmol/L, and/or the presence of an anti-diabetic drug as current treatment for diabetes mellitus [9]. We included patients presenting to or patients in our hospital who were found to have a moderate or severe DFUI, based on the criteria developed for IDSA foot infection guidelines [10]. Briefly, a diabetic foot infection is defined as having ≥2 manifestations of inflammation (local swelling or induration, erythema, tenderness or pain, warmth, or purulent discharge (thick, opaque to white or sanguineous secretion). Moderate infection was defined as local DFUI with surrounding erythema > 2 cm, or involving structures deeper than skin and subcutaneous tissues (e.g., abscess, osteomyelitis, septic arthritis, fasciitis) but no evidence of systemic inflammatory response syndrome. Severe infection was defined as local DFUI accompanied by ≥2 systemic signs of inflammatory response syndrome. Cure was defined as absence of microbiological, clinical, and imaging evidence of the original infection. Improvement was defined as a significantly better wound score and a decrease in the number of manifestations of inflammation, however without complete resolution. Regarding the evolution of the wound, we used a custom-made DFUI assessment to describe the evolution of the wound, summing up various elements of a wound in a single wound score (Additional file 1). The elements used for that score were all quantified, and consisted of the wound size, wound depth, the degree of undermining of wound edges, the presence of discharge and pus, erythema, induration, tenderness, pain and local warmth (Additional file 1).
A patient was eligible for study participation if he/she met the following criteria: is aged ≥18; has been diagnosed with diabetes mellitus; has an open foot wound of ≥1 cm2 located on or below the malleolus that has findings of infection; has undergone any appropriate surgical intervention needed to remove necrotic and to drain an abscess (Fig. 1); and, if female, is nonpregnant and nonlactating. Of note, surgical debridement was not a must. Patients could participate in the study without prior surgical debridement if there was no clinical need to remove necrosis or to drain an abscess. The exclusion criteria were: a known history of hypersensitivity to gentamicin or bovine collagen [11]; a DFUI associated with prosthetic material or any implanted device; peripheral arterial insufficiency clinically requiring urgent limb revascularization after inclusion; and, having received > 48 h of potentially effective antibiotic therapy and showing improvement in wound infection. Of note, a chronic arterial insufficiency with a clinically preserved foot in absence of vascular necroses was not an exclusion criterion. If a patient had received an antibiotic within 72 h, but deep-tissue culture results indicated that the infecting pathogen was not susceptible to that antibiotic, the patient could be enrolled. Further exclusion criteria were: mild DFUI (as defined by the IDSA criteria [10]); severe immune-suppression; extensive necrosis requiring amputation (Fig. 2); residual osteomyelitis (after any debridement) requiring more than 28 days of antibiotic treatment [1]; a history of myasthenia gravis or other neurological conditions precluding gentamicin use; a history of epilepsy; and, active or recent alcohol or substance abuse.
Gentamicin-collagen sponges
Innocoll Pharmaceuticals Ltd. provided us with the bioresorbable gentamicin-collagen sponges. These are 5 × 5 cm in size and contain type 1 bovine collagen and 50 mg of gentamicin sulfate (equivalent to 32.5 mg of gentamicin base), and are commercially available in Switzerland under the trade name GARAMYCIN® Sponge (Fig. 3; unpacked sponge). We made an educational videotape to demonstrate the correct application of the sponge (Additional file 2: Video S1).
Study conduct
This was a prospective, randomized, investigator-blinded, controlled, single-centre study. Using a blinded allocation scheme, patients were randomized 1:1 to receive systemic antibiotic therapy and standard ulcer care, with either (A) daily application of a gentamicin-sponge, or (B) no-sponge. The clinical investigator (IU or BK) selected the highest severity ulcer (if there was more than one) and determined the number of sponges (up to 4) that a patient required to fully cover the wound. If needed, the patient underwent surgical debridement (or partial amputation) as long as there was residual infected wound. The patient could also undergo limb revascularization if the procedure was performed before they started in the study. We did not impose any study-related conditions regarding the surgical or angiologic techniques provided to the patient.
We allowed enrolled patients to have been on an effective systemic antibiotic regimen for up to 48 h prior to study inclusion. Once the patient was enrolled in the study, the clinical investigator (IU, an infectious diseases physician) prescribed a systemic antibiotic regimen (empiric if there were no culture and sensitivity results, specific if there were), based on a scheme developed specifically for this protocol. For both study arms, the preferred systemic (oral or intravenous) antibiotic treatment was levofloxacin (with or without additional clindamycin if necessary), or amoxicillin-clavulanate. In cases of severe infection or sepsis, we treated the patient with piperacillin-tazobactam or aztreonam, if needed [10]. If infection with an obligate anaerobe was clinically suspected or confirmed [12], we added metronidazole (if the patient was not already receiving clindamycin). If cases with suspected or proven infection with methicillin-resistant staphylococci [13], we added linezolid to the regimen. Standard wound care included sharp debridement (at enrollment and during hospitalization or at clinic visits, if clinically indicated), dressing changes (moisture of 0.9% saline for those not treated with the gentamicin-collagen sponge), pressure off-loading and efforts to optimize glycemic control. Enrolled patients could not be treated with topical antiseptics or antibiotics (other than the gentamicin-sponge), hyperbaric oxygen therapy or vacuum-assisted negative-pressure devices.
Patients were scheduled to receive approximately 14 to 28 days of study treatment. They were seen for follow-up in the hospital (if an inpatient) or if treated as outpatients, in the clinic, weekly (on nominal days 8, 15, 22, and 29) for safety and efficacy assessments. The end-of-treatment visit was defined as the patient’s last visit on which study treatment was formally discontinued for any reason. The final assessment was performed at the test-of-cure visit, approximately 10 days after treatment was discontinued. For each enrolled patient we assessed variables pertaining to demographic characteristics, immune-suppression status, wound microbiology results, surgical procedures, and antibiotic treatment. A microbiological culture of tissue collected from the target ulcer (obtained by debridement, curettage, or biopsy, but not swab) was performed at baseline and at the final visit (if there was ulcers left for debridement). We noted eradication or persistence of baseline pathogens, based on culture results reported by the local laboratory. We incubated all specimens for up to 5 days and processed any isolates according to CLSI recommendations [14], before switching to EUCAST criteria (European Committee) in 2014 [15]. Finally, we made safety assessments at each visit during the study. Investigators specifically inquired about any vestibular disturbances and ototoxicity, such as nausea, tinnitus, vertigo, and decrease in hearing.
A research nurse (BK) and an infectious diseases physician (IU), who both specialized in DFUI, supervised enrollment and follow-up of all patients, distinguished between infection and colonization and between causative pathogens and less virulent microorganisms. A second research nurse (LH) monitored the study. Innocoll Ltd. made a donation to our medical center to support the costs of conducting this pilot study. This donation did not come with any conditions or requirements regarding the academic freedom and scientific publications. We registered the study on
ClinicalTrials.gov
(NCT01951768) on 2 April 2013.
Statistical analyses
Our primary objective was to determine the effect of the topical gentamicin-collagen sponge in combination with systemic antibiotic therapy compared to systemic antibiotic therapy alone on clinical outcome in the treatment of a moderate or severe DFUI. Secondary objectives were: to determine the effect of the gentamicin-sponge on the eradication of baseline ulcer pathogens; to assess the safety and tolerability of the gentamicin-sponge; and, to determine the rapidity of wound healing over time. Assuming a clinical “clinical cure” in 90% of patients treated with the gentamicin-sponge versus 70% in the control group (no-sponge), we aimed for a sample size of 144 subjects to all 80% power to detect superiority of the gentamicin-sponge arm at the 0.05 significance level.
We performed group comparisons using the Pearson-χ2-test, the Fisher-exact, or the Wilcoxon-ranksum-test, as appropriate. We performed an unmatched logistic regression analysis with the outcome “cure”, adjusted for the case mix. We introduced independent variables with a p value ≤0.05 in univariate analysis stepwise into the multivariate analysis, except for variables for gentamicin sponges, which we automatically included in the model. We included 12 predictor variables per outcome [16] and checked variables for collinearity and interaction. We used STATA software (9.0, STATA™, USA). P values ≤0.05 (two-tailed) were significant.