Impact of wound microbiology on limb preservation in patients with diabetic foot infection

Abstract Aims/Introduction: To investigate the association between specific bacterial pathogens and treatment outcome in patients with limb‐threatening diabetic foot infection (LT‐DFI). Materials and Methods: Consecutive patients treated for LT‐DFI in a major diabetic foot center in Taiwan were analyzed between the years 2014 and 2017. Patients with positive wound culture results at first aid were enrolled. Clinical factors, laboratory data, and wound culture results were compared. Lower‐extremity amputations and in‐hospital mortality were defined as a poor outcome. Results: Among the 558 patients, 272 (48.7%) patients had lower extremity amputation and 22 (3.9%) patients had in‐hospital mortality. Gram‐negative bacterial (GNB) infection was the independent factor following factors adjustment. When all the 31 microorganisms were analyzed, only E. coli (adjusted odds ratio [aOR], 3.01; 95% CI, 1.60–5.65), Proteus spp. (aOR, 2.99; 95% CI, 1.69–5.29), and Pseudomonas aeruginosa (aOR, 2.00; 95% CI 1.20–3.32) were associated with poor outcome. The analysis of specific GNB species in association with major‐ or minor‐ amputation have been reported. No specific pathogen was associated with cause of death in patients with mortality within 30 days. The antimicrobial‐resistant strains were not associated with a poor treatment outcome. Conclusions: The presence of GNB was associated with limb amputations. This study provides insight into more timely and appropriate management of the diabetic foot infection.


INTRODUCTION
Diabetic foot ulcer (DFU) is a major complication of diabetes and contributes to most causes of non-traumatic lowerextremity amputation (LEA) [1][2][3][4] . Among factors such as perfusion, wound depth and width, infection and neuropathy that might affect wound severity, diabetic foot infection (DFI) is one of the major factors attributed to limb loss while treating DFU [5][6][7][8] .
According to the Infectious Diseases Society of America (IDSA) and the International Working Group on Diabetic Foot (IWGDF) guidelines, DFI has been classified into four grades 9 .
Patients with severe foot infection (grade 4) and those with moderate infection (grade 3) plus relevant morbidities are under limb-threatening DFI (LT-DFI) status and usually need interdisciplinary limb management 9,10 . In our recent study 7 , major adverse cardiovascular event (MACE), poor peripheral circulation as well as the grade of DFI were associated with these poor prognoses (LEA and in-hospital mortality).
Previously, gram-positive aerobic bacteria have been reported to be the dominant pathogen in mild to moderate DFI by tissue curettage 11 . Staphylococcus aureus is common in patients with acute DFI, whereas gram-negative bacteria and obligatory anaerobic bacteria are more present in chronic foot infection 12 .
Pseudomonas aeruginosa causes DFI more often in hot and humid areas 13 . However, to our best knowledge, the impact of microorganisms in patients with LT-DFI has not been well recognized. This study aimed to further analyze the bacteria in patients with LT-DFI (grades 3 and 4) in order to research the association between specific pathogens and treatment outcome.

Data source and study population
This study used an anonymous public use data set without identifiable information about individuals in the study. It was approved by the Institutional Review Board of Chang Gung Memorial Hospital (No. NMRPG3K0391), and follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.
From January 1, 2014 to December 31, 2017, 722 type 2 diabetic patients with LT-DFI (infection grade ≥ 3) were consecutively reviewed and treated in the diabetic foot care unit at Chang Gung Memorial Hospital, an interdisciplinary diabetic foot care center accredited by the International Diabetes Federation West-Pacific Region in Taiwan 14 . Every patient received wound culture on admission. Patients with a negative culture result (n = 164) were excluded, while 558 patients with documented culture results were enrolled in this study.

Wound recording
Wound classification was recorded as PEDIS describing the perfusion, extent size, depth/tissue loss, infection, and sensation of the wounds 15 . LT-DFI was defined by an infection score greater than 3. Patients with grade 3 infection presented with either erythema >2 cm around the wound or the structures deeper than skin and subcutaneous tissues were involved, but with no systemic inflammatory response syndrome (SIRS). The definition of SIRS was according to matching two or more of the four criteria including abnormal body temperature >38°C or <36°C; tachycardia with pulse >90 beat per minute; abnormal respiratory rate with >20 breaths per minute; and an abnormal leukocyte count >12,000 or <4,000 /cu mm.
The perfusion status was categorized into three grades, with grade 3 perfusion status representing critical limb ischemia as defined by the presence of gangrene or ulcers with an ankle pressure <70 mmHg 16 , or monophasic wave form of distal segment of the posterior tibial artery and dorsalis pedis artery 17 . Adjunct angiography was performed for confirmation.

Wound culture
For clinically infected ulcers, we obtained specimens by cotton swabs (Transystem, COPAN, Italia) for both bacteria and fungus culture after necrotic tissue and surrounding callus debridement at first aid was performed to determine the causative pathogens. Ninety-five percent of specimens in this study were obtained from deep ulcers (penetrating to fascia, muscle, tendon, and/or bone). Culture media included chocolate agar, sheep blood agar, and thioglycolate broth at 37°C. Mycosel agar plates were also obtained and maintained at 25°C to enhance fungal growth. Positive microbial cultures were defined as growth of the same pathogen on two or more culture media with positive fungal culture defined on morphology. Parenteral broad-spectrum antibiotic therapy was initiated empirically for common gram-positive and gram-negative bacteria, and obligatory anaerobic pathogens with antibiotic regimen were adjusted based on both the clinical response to empirical therapy and sensitivity results. Cultures were repeated for patients who were not responding to appropriate therapy.
Antibiotics strategy and consensus of managements Broad-spectrum antibiotics against gram-negative and anaerobic pathogens, including third-generation cephalosporin (43%), extended-spectrum penicillin (31% with aminopenicillins and 5% with ureidopenicillins), fluoroquinolones (6%), carbapenems (5%), and metronidazole (19%) were prescribed promptly for these patients initially. Glycopeptide against methicillin-resistant Staphylococcus aureus (MRSA) was also prescribed in 26% of patients. Empiric antibiotics were subsequently modified according to the results of cultures. Surgical interventions, endovascular treatments, and foot amputations were scheduled in a timely manner after the diabetic foot team reached consensus.

Outcomes definitions
Those patients receiving minor (below the ankle) or major (above the ankle) amputations or expiring during the hospital treatment course were defined as poor outcomes.
Since various causes of death including nosocomial infection of patients with DFU were noted for a longer stay of hospitalization 14 , only those patients who died within 30 days of hospitalization were analyzed to find the association between pathogens and mortality.

Statistical analysis
Clinical demographics, associated comorbidities (such as hypertension, history of MACE, and dialysis), and factors of PEDIS wound-grading were recorded from the patient's first visit at admission, with laboratory data of routine hematology tests and chemistry profile at enrolment being analyzed. Categorical variables were reported as numbers with percentages, and continuous variables were reported as means and standard deviations. Comparisons between patients with preserved limbs or poor outcome were performed using the Mann-Whitney test for continuous variables and Pearson's chi-square test for categorical variables. Each factor odds ratio to poor outcome was calculated via an adjusted model of logistic regression, while the same statistical method was used in comparing the two groups with different treatment outcomes. The significant risk factors in the univariate analysis found above were then entered into a multivariate logistic regression model to identify independent risk factors to adverse outcome among these patients. All statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS for Mac, version 26.0, IBM Corp., Armonk, NY, USA) data analysis software. 7% of grade 2 sensation, P < 0.001). The incidence of osteomyelitis was higher in patients with poor outcome (51.4% vs 40.0%, P = 0.017).
A higher leukocyte count (15.2 vs 12.3 10 3 /mL, P < 0.001) and c-reactive protein level (131.3 -99.9 vs 90.3 -89.3mg/dL, P < 0.001) were both found in the poor outcome group, indicating the severity of infection status, while lower hemoglobin, serum albumin, and high-density lipoprotein were also revealed. There were no statistical differences in HbA1c. Of note, higher incidences of gram-negative and obligatory anaerobic bacteria were documented in the poor outcome group.  1%). When putting baseline characteristics with statistical significance into the multivariate logistic regression model, gram-negative aerobic bacteria (odds ratio 2.59, 95% confidence interval 1.77-3.79) and obligatory anaerobic bacteria (OR 2.28, 95% CI 1.58-3.29) infection could predict poor outcome. We further adjusted wound assessment, laboratory data with statistical significance, and the incidence of osteomyelitis, but gram-negative bacteria remained the independent factor for poor outcome in these groups ( Table 2).
Forest plot analysis and unadjusted odds ratio of specific pathogens revealed poor outcome in patients with LT-DFI as demonstrated in Figure 1  After adjustment of baseline characteristics and laboratory data with statistical significance and wound assessment, Proteus spp. and Klebsiella spp. remained statistically significant in minor and major amputations, respectively.

Prevalence of antimicrobial resistance strains
In order to investigate the role of antimicrobial-resistant strains, the cultured results of the causative pathogens in poor treatment outcome were further analyzed (Table 3). Ceftriaxone-or ceftazidime-resistant strains accounted for 2.7% (1/36) in Klebsiella spp., while low percentages of carbapenem-resistant strains (5.6% and 1.6% in Klebsiella spp. and E. coli, respectively) were found.

DISCUSSION
In addition to the well-known factors of poor outcome (e.g. MACE, ESRD, PAD, large wound size, deep wound, impaired sensation, severe infection, and the presence of osteomyelitis), our results demonstrated the association between gram-negative bacteria and poor in-hospital treatment outcome. After multivariate multinomial logistic regression analyses of these clinical factors with statistical significance, gram-negative bacteria remained an independent factor for poor outcome in our referral center. Among these, Proteus spp. and Klebsiella spp. could predict minor and major amputations, respectively, even with low percentages of antimicrobial resistance strain.
DFI is a heterogenous group of infection in a diverse patient population 18 , with the causative pathogens of DFI varying by geographic, demographic, and clinical situations. Gram-positive bacteria, especially S. aureus, are reported to be the predominant and virulent pathogens in mild or moderate DFI 11,19 ; however, the incidence of gram-negative bacteria infection has become more common in warmer climates 20 and chronic wounds 21 . Compared with gram-positive infection, gramnegative bacteria induce a greater magnitude of inflammatory response 22 , leading to LEA 23 and high in-hospital mortality 24 . In the diabetic foot, gram-negative bacterial infection has been reported to induce ischemic tissue necrosis and thereby progressive infection and poor outcome 25,26 . Thus, infections with gram-negative bacteria might generally require more frequent surgical debridement, endovascular therapy, and longer hospitalization than those with gram-positive bacteria.
The predominance of the Enterobacteriaceae family (E. coli, Klebsiella spp., Morganella morganii, and Proteus spp.) has recently been reported as the largest group of aerobic gramnegative bacteria in DFI 27 . Extraintestinal pathogenic E. coli is well-known for its virulence potential to invade host tissue and to be transmitted via the bloodstream 28 . Klebsiella spp. can cause severe infection with a high mortality rate 29 , is a common pathogen of pyogenic liver abscess 30 , and potentially fatal necrotizing fasciitis in Taiwan 31 . Proteus spp. exhibits a characteristic swarming activity, which enables colonization, tissue invasion, and biofilm formation in chronic wounds 32 , and is one of the most common pathogens of osteomyelitis 33 , and diabetic patients with osteomyelitis are prone to limb loss if this is accompanied by exposed bone, the presence of ischemia, and necrotizing soft tissue infection 34 . Major LEA is indicated in these patients when accompanied by severe sepsis, extensive tissue loss, and poor-healing wound(s). This study demonstrates the importance of specific pathogens in the Enterobacteriaceae family and their role in amputations and in-hospital mortality.
In the past few years, antibiotic-resistant pathogens, particularly MRSA and extended-spectrum b-lactamase-(ESBL) or carbapenemase-producing gram-negative bacteria have become a major problem in treating DFI 35 . DFIs caused by MRSA have been thought to have worse outcomes; however, a recent review  37 . In our study, ceftriaxone-or ceftazidimeresistant strains were low in Klebsiella spp. and E. coli, indicating that a poor outcome was not associated with antibioticresistant strains.
The main limitation of this study is its retrospective nature. In addition, wound culture samples in patients with LT-DFI were obtained by swab, the most common real-world practice. Previous broad-spectrum antibiotic treatment before patient referral might influence the culture result on admission. Although new advances obtained from DNA-and RNA-based techniques for bacterial identification could improve therapeutic approaches 27

CONCLUSION
This study demonstrated that the presence of gram-negative bacteria ought to raise awareness in clinicians treating patients with LT-DFI. Further evaluation of these specific species might provide further insight for the management of DFI.

FUNDING SOURCES
The study was supported by project grant from the Ministry of Science and Technology, Taiwan (MOST 109-2314-B-182A-128).

DISCLOSURE
The author declares no conflict of interest.