Ethical approval for this study was granted by the Ethics and Research Committees of the Obafemi Awolowo University Teaching Hospitals Complex and Ladoke Akintola University College of Technology with protocol numbers ERC/2015/11/02 and LTH/ER/2016/01/254. Information about the study and participant involvement was fully explained to patients, and properly signed and dated written informed consent forms were obtained from patients before their recruitment into the study. Results of wound biopsy microscopy, culture and sensitivity were made available for patients’ management.

The prospective, cross-sectional, multicentre, hospital-based study was carried out in Osun state, southwest Nigeria, between July 2016 and April 2017. It included three existing tertiary healthcare facilities in the state: Obafemi Awolowo University Teaching Hospitals Complex, Wesley Guild Hospital, Ilesa, and Ladoke Akintola University of Technology Teaching Hospital, Osogbo. All consecutive diabetic patients (both hospitalised and outpatients) with foot infections meeting the criteria for diagnosis of IDFU seen and managed at these hospitals were recruited into the study. They were clinically assessed and foot lesions graded according to the diabetic foot infection severity classification system issued by the Infectious Disease Society of America.⁸ Only non-duplicate patients and samples were included in the study. All inpatients were followed up with regular check-ups physically in the wards until they either died or were discharged.

Aspirates were obtained from deep-seated abscesses, and tissue samples were collected after washing the wound vigorously with sterile saline and debridement of the slough to exclude mere colonisers. Necrotic tissues were curetted into Anaerobic Basal Broth (Oxoid, Basingstoke, Hants, United Kingdom) for anaerobic culture. The samples were immediately transported to the laboratory and processed within 2 h of sample collection by inoculating them onto a set of selective and non-selective media which were: 5% (volume/volume) sheep blood agar (BA; Oxoid, Basingstoke, Hants, United Kingdom), MacConkey agar (Oxoid, Basingstoke, Hants, United Kingdom), chocolate agar and anaerobic basal agar (Oxoid, Basingstoke, Hants, United Kingdom) supplemented with 5% (abscesses) laked sheep blood, Vitamin K1 (1 µg/mL), L-cysteine hydrochloride (5 µg/mL) and gentamicin (100 µg/mL) (gentamicin blood agar).

Inoculated plain BA and MacConkey agar plates were incubated in air and chocolate agar plates in CO₂ at 37 °C for 24 h. Inoculated plain gentamicin BA plates, as well as gentamicin BA with kanamycin (75 µg/L) and vancomycin (5 µg/L) supplements, were incubated under anaerobic conditions made up of 80% H₂, 10% CO₂ and 10% N₂ for 48 h and extended for 5 days if necessary; anaerobiosis was achieved using a Bactron anaerobic chamber (Sheldon Manufacturing, Inc., Cornelius, Oregon, United States). Representative colonies were identified by colonial morphology, Gram staining characteristics and conventional biochemical tests including catalase and oxidase tests. Facultative anaerobic Gram-negative bacilli (GNB) and Streptococcus spp. were further identified with Microbact^™ GNB 24E (Oxoid, Basingstoke, Hants, United Kingdom) and RapID^™ STR (Thermo Fisher Scientific, Remel Products, Lexena, Kansas, United States), while Staphylococcus spp. were further identified with a coagulase test, characteristic growth appearance on mannitol salt agar and a DNAse test. The obligate anaerobes were identified by RapID^™ ANA II (Thermo Fisher Scientific, Remel Products, Lexena, Kansas, United States). Yeast isolates were identified by Gram staining and germ tube tests. Quality control strains, Staphylococcus aureus ATCC 25923, Escherichia coli ATCC 25922, Bacteroides fragilis ATCC 25285 and Peptostreptococcus anaerobius ATCC 27337, were used to assess the quality of the media and identification systems. The quality of our bacterial identification system and procedures (for aerobes, facultative anaerobes and obligate anaerobes) were assured by ensuring that the control bacterial strains were identified to their names.

Antibiotic susceptibility testing for aerobes and facultative anaerobes was performed using the modified Kirby-Bauer disc diffusion technique as recommended by the Clinical and Laboratory Standards Institute (CLSI).²¹ Gram-positive isolates were tested with the following antibiotic discs (Oxoid, Basingstoke, Hants, United Kingdom): penicillin (10 µg), cefuroxime (30 µg), ceftriaxone (30 µg), cefepime (30 µg), co-amoxiclav (20/10 µg), ciprofloxacin (5 µg), chloramphenicol (30 µg), gentamicin (10 µg), amikacin (10 µg), trimethoprim-sulfamethoxazole (1.25 µg/23.75 µg), cefoxitin (30 µg), erythromycin (15 µg), ampicillin-sulbactam (10 µg/10 µg), piperacillin-tazobactam (100 µg/10 µg) and vancomycin (30 µg). Vancomycin (256–0.015 µg/mL) minimum inhibitory concentration (MIC) strip (Oxoid, Basingstoke, Hants, United Kingdom) was used to test for vancomycin resistance among methicillin-resistant S. aureus (MRSA). Gram-negative isolates were tested with the following antibiotic discs (Oxoid, Basingstoke, Hants, United Kingdom): ceftriaxone (30 µg), ceftazidime (30 µg), cefotaxime (30 µg), cefepime (30 µg), co-amoxiclav (20/10 µg), ciprofloxacin (5 µg), gentamicin (10 µg), amikacin (10 µg), chloramphenicol (30 µg), aztreonam (30 µg), trimethoprim-sulfamethoxazole (1.25 µg/23.75 µg), ampicillin-sulbactam (10 µg/10 µg), piperacillin-tazobactam (100 µg/10 µg), cefoxitin (30 µg), meropenem (10 µg) and ertapenem (10 µg).

Discrete colonies were emulsified in sterile saline to match 0.5 McFarland turbidity standards from where confluence inocula were made on Mueller-Hinton agar (MHA) plates with sterile cotton swabs. The swabbed MHA plates were allowed to dry at room temperature and a set of six antibiotic discs were placed evenly spaced on each of the plates. Vancomycin resistance was tested in the methicillin-resistant S. aureus isolates by placing a MIC evaluation strip (Oxoid, Basingstoke, Hants, United Kingdom) on inoculated MHA. After 18–24 h of incubation, the diameter of the zone of inhibition around each antibiotic disc was measured and recorded. Vancomycin MIC values were also recorded for the S. aureus. Zones of inhibition of each antibiotic as well as vancomycin MIC values were interpreted as ‘sensitive’, ‘intermediate’ or ‘resistant’ in accordance with CLSI guidelines.²¹ Isolates with intermediate sensitivity were regarded as ‘resistant’. The quality of antibiotic susceptibility testing consumables (including antibiotic discs and MHA) and procedures were assured with bacterial control strains (E. coli ATCC 25922, S. aureus ATCC 25923 and S. aureus ATCC 43300). Zones of inhibition of tested antibiotics on the bacterial control strains fell within their quality control ranges according to CLSI.²¹

Extended-spectrum β-lactamase production was confirmed among Enterobacteriaceae and other GNB that showed reduced susceptibility to at least one of the tested third-generation cephalosporins (cefotaxime 30 µg, ceftazidime 30 µg and ceftriaxone 30 µg) or aztreonam (30 µg) by a combination disc diffusion test (CDDT).²¹ CDDT was done using both single discs of cefotaxime (30 µg) and ceftazidime (30 µg) and their respective clavulanic acid containing discs (cetotaxime-clavulanate 30/10 µg, ceftazidime-clavulanate 30/10 µg). A 5 mm or more increase in zone of inhibition of one or more combination discs as compared with their respective single discs was taken as confirmatory evidence of extended-spectrum beta-lactamase (ESBL) production.²¹

AmpC beta-lactamase production was detected by AmpC disc test as described by Anjali et al. on isolates which show resistance to at least one third-generation cephalosporin and a β-lactamase inhibitor.²² A broth culture of E. coli ATCC 25922 was adjusted to 0.5 McFarland turbidity standard and inoculated onto MHA plates. Sterile filter paper discs (6 mm) were moistened with distilled water (about 20 µl) and up to five colonies of the test organism was transferred onto the filter paper. Afterwards, a cefoxitin (30 µg) disc and the inoculated filter paper disc were placed next to each other and almost touching on inoculated media. This setup was incubated overnight at 37 °C. A flattening or indentation of the zone of inhibition of cefoxitin in the vicinity of the test disc (inoculated filter paper) indicated a phenotypic confirmatory evidence of AmpC β-lactamase production.²²

Gram-negative bacilli with intermediate sensitivity or resistance to one or more carbapenems were tested for production of carbapenemases by the modified Hodge test and interpreted by CLSI guidelines.²¹ Organisms that were phenotypically MDR, including ESBL-producing GNB, carbapenem-resistant GNB and MRSA, were further tested for resistance-determining genes using polymerase chain reaction (PCR)-based protocols with specific oligonucleotide primers^23–27 (Table 1); template bacterial DNA was extracted by the boiling method.²⁸ Electrophoresis of each PCR product (5 µL) was carried out in 1.5% (weight/volume) Agarose gel (Biomatik, Kitchener, Ontario, Canada) in 1X Tris-Acetate-EDTA (TAE) buffer for 45 min. The size of amplified products was estimated using 100 base pairs molecular weight marker (100–1200 base pairs).

Data analysis was performed with Statistical Package for Social Sciences version 20 (SPSS Inc., Chicago, Illinois, United States). Comparison of mean values was done using the Student’s t-test for continuous variables and the chi-square test for categorical variables. Risk factors for infection of diabetic foot by MDR organisms were identified by logistic regression analysis. Logistic regression was used to determine predictive associations of variables that showed statistical significance by bivariate analysis. A p-value of less than 0.05 was considered to be statistically significant.

Ninety patients (53 male and 37 female) presented with 93 IDFUs during the 11-month study period. The patients ranged between 18 and 85 years (mean 54.7 ± 12.8 years) of age. Of the 93 cases of foot infections, 56 (60.2%) had lasted for at least 1 month, and 70 (75.3%) were in-hospital patients. Sixty-six (71.0%) of the ulcers were categorised as Wagner’s grade 3 and above. Most (n = 74; 82.2%) of the patients used antibiotics in the month preceding presentation at the healthcare facilities, while 93.3% (n = 84) had commenced antibiotics before collection of wound samples (Table 2).

Significant factors associated with the presence of MDR organisms in diabetic foot infections included peripheral sensory neuropathy, a foot infection duration of more than a month and admission duration of more than a month (Table 2). Further analysis with logistic regression however identified only peripheral neuropathy (adjusted odds ratio = 4.05, 95% CI: 1.08–15.13) and foot infection duration of more than a month (adjusted odds ratio = 7.63, CI: 1.64–35.39) as the predisposing factors for acquisition of multidrug-resistant bacteria among patients with IDFU. A substantial proportion (33/70; 47.1%) of the inpatients had poor treatment outcomes; poor outcomes noted in 53.4% (31/58) of patients with MDR infections included major limb amputation (below and above knee amputation) (18/58; 31.0%) and death (13/58; 22.4%). Infections by these MDR bacteria have significant association with poor treatment outcomes (adjusted odds ratio = 5.11, 95% CI: 1.23–29.67).

All 93 wound specimens obtained from the 90 patients (three patients had bilateral foot ulcers) in the three healthcare facilities were positive on bacterial culture with only 10 (10.8%) of them yielding a single organism each. Among the polymicrobial cultures, 45 (48.4%) yielded two organisms per culture, 34 (36.6%) yielded three organisms per culture while 4 (4.3%) yielded four organisms per culture. Results further showed that there was a total of 218 organisms isolated from the 93 specimens cultured with an average of 2.34 organisms per sample. Of the organisms, 129 (59.2%) were Gram-negative aerobic and facultative anaerobic bacilli, 59 (27.1%) were Gram-positive aerobic cocci and 29 (13.3%) were anaerobic bacteria; only 1 (0.5%) was yeast. S. aureus (34; 15.6%) was the single most common organism followed by E. coli (23; 10.6%) and Pseudomonas aeruginosa (20; 9.2%). Others included Klebsiella spp. (19; 8.7%), Citrobacter spp. (19; 8.7%), Enterococcus spp. (14; 6.4%), Enterobacter spp. (11; 5.0%), Proteus mirabilis (10; 4.6%) and Acinetobacter spp (9; 4.1%). On the other hand, the predominant anaerobic bacteria were Bacteroides spp. (7; 3.2%) and P. anaerobius (6; 2.8%) as shown in Table 3.

Gram-positive bacteria were highly resistant to trimethoprim-sulfamethoxazole (69.5%), penicillin G (66.1%) and gentamicin (40.1%) but demonstrated low-level resistance to piperacillin/tazobactam (6.8%) and amikacin (10.2%). On the other hand, Gram-negative bacteria were highly resistant to the third-generation cephalosporins which included ceftriaxone (56%), cefotaxime (55%) and ceftazidime (48.1%), as well as trimethoprim-sulfamethoxazole (89%), gentamicin (54.3%) and ciprofloxacin (54.3%). Low rates of resistance were however shown to ertapenem (6.4%), piperacillin/tazobactam (9.3%) and amikacin (12.4%).

Further analysis of resistance profiles in the organisms showed that of the 188 aerobic isolates, 121 (64.4%) were MDR, being resistant to one or more agents in at least three antibiotic classes (Table 4). The prevalence rates of MDR among GPC and GNB were 55.9% and 68.2%. Multidrug resistance rates were generally high among the isolated bacteria especially Acinetobacter species (88.9%), Enterococcus species (84.6%), Enterobacter species (81.8%) and Citrobacter species (73.7%). Overall prevalence of MDR bacteria among the IDFU cases was 80% (n = 74) with rates among in-patient and outpatient cases being 82.9% (n = 58) and 69.6% (n = 16). Twelve (35.3%) S. aureus were methicillin-resistant, while of the 129 GNB, 43 (33.3%) were ESBL-producing and 10 (7.8%) were carbapenem-resistant (Table 5). High ESBL production rates were seen among Enterobacter species (54.6%), Klebsiella species (52.6%), Citrobacter species (52.6%) and E. coli (43.5%). Other ESBL-producing species found were Hafnia alvei (1; 20.0%), Providencia spp. (1; 33.0%) and Morganella morganii (2; 28.6%). AmpC β-lactamase production was detected among Citrobacter spp. (1; 5.3%), E. coli (2; 8.7%) and M. morganii (1; 14.3%). Carbapenem resistance was seen among Acinetobacter baumannii (4; 44.4%), H. alvei (2; 40.0%), P. aeruginosa (3; 15.0%) and M. morganii (1; 14.3%). Further tests showed that among the 10 carbapenem-resistant isolates, only four, H. alvei (2/4) and A. baumannii (2/4), were carbapenemase-producing.

Ten (83.3%) of the 12 MRSA isolates harboured the mecA gene. At least one of the ESBL genes investigated was detected in 86.0% (n = 37) of the 43 ESBL-producing organisms. The most common ESBL gene detected was bla_CTX-M, harboured by 30 (69.8%) of the phenotypically confirmed ESBL-producing isolates. Others were bla_TEM (27; 62.8%) and bla_SHV (8; 18.6%) (Table 6). Thirty-one (72.1%) of the ESBL-producing GNB had at least two ESBL genes. Four of the 10 carbapenem-resistant species were phenotypically confirmed to be carbapenemase-producers. Carbapenemase and metallo-beta-lactamase genes were detected in all of the four phenotypically confirmed carbapenemase-producers; they were bla_VIM (n = 3), bla_KPC (n = 2) and bla_NDM (n = 1). These genes were detected in all of the carbapenemase-producing H. alvei (2/2) and A. baumannii (2/2).

The limitation of the study was that the number of patients recruited was limited to 90 and this was because the study was time-bound. Also, outpatients could not be followed up because of the multicentre nature of the study. Resistance profiles of obligate anaerobic bacteria could not be determined and whole genome sequencing (for strain relatedness) was also not done due to lack of financial support for the study.

The spectrum of agents causing IDFU is wide and includes numerous species of aerobic and anaerobic bacteria. There is a high prevalence of MDR aerobic bacteria among them which poses a great limitation to the effective treatment of cases.