Prescribing antibiotics in diabetic foot infection: what is the role of initial microscopy and culture of tissue samples?^*

Abstract

The aim of this study was to evaluate the role of microscopy, Gram stain and the culture of tissue samples in the antibiotic treatment of patients with diabetic foot infection. A retrospective review of patients with a diabetic foot infection was undertaken. Data analysed included the severity of infection, antibiotic prescribing patterns, microscopy and culture results. A total of 71 patients were included, from whom 114 tissue samples were collected. Gram stain results were in agreement with final culture results in 45·8% (n = 54) of samples. Overall sensitivity and specificity of the Gram stains were low (74·5% and 69·8%, respectively), although the specificity for Gram‐negative rods was high (98·5%). The presence or absence of ‘pus cells’ on microscopy was a poor predictor of culture results. Empirical prescribing of antibiotics was in accordance with local policy in 31·1% of patients, improving to 86·8 % following culture results. Microscopy, a skilled laboratory procedure, was generally a poor predictor of tissue culture results. However, the presence of Gram‐negative rods was suggestive of isolation in the culture of such organisms and could allow the early broadening of antibiotic treatment. Despite initial poor compliance of empirical antibiotic treatment regimens, prescribing was adjusted in light of culture results, suggesting these were important for clinicians.

Introduction

Foot ulceration with subsequent infection is a frequent complication of diabetes 1. As the worldwide prevalence of diabetes increases, the lifetime incidence of a foot ulcer within this population is estimated to be as high as 25% 2, 3. The newly updated International Working Group for Diabetic Foot (IWGDF) guidelines on foot infection recommend the use of the Infectious Disease Society of America (IDSA) classification system to establish the clinical severity of any diabetic foot infection (DFI). In order to provide targeted therapy, a consensus recommendation amongst these and other guidelines is to collect a wound specimen before commencement of empirical antibiotics to establish the causative organism(s) and antibiotic sensitivities 1, 4, 5.

Before culture results are available, the tissue sample Gram stain can provide an early indication of an infection through the presence of microorganisms or polymorphonuclear leukocytes (‘pus cells’) 1, 4. The test results can be available in ‘real time’ (about 30 minutes); it is cheap, easy to perform and potentially gives the clinician valuable information about the likely causative pathogens. Currently, there is little published data on the value of these early tests in the management of DFI. In addition, there is also a lack of evidence for whether clinicians are using the microbiological culture data to guide the antibiotic regimens.

The aim of this study was to evaluate the performance of microscopy and Gram staining in predicting the culture results in DFI and to examine whether microbiological results are used by clinicians to aid their prescribing.

Materials and methods

Queen Elizabeth Hospital Birmingham is a UK 1200‐bed hospital, providing both secondary and tertiary care services. The diabetic foot service is a multidisciplinary team, including a diabetologist, vascular surgeon, microbiologist and podiatrists, reviewing approximately 350 outpatients per year. All patients who had a diabetic foot tissue sample taken between January 2012 and December 2013 were identified retrospectively from the Microbiology laboratory system (Telepath) database, and data were gathered on the microscopy, Gram stain, culture and antibiotic sensitivity results.

Further information was collected from the hospital electronic patient record on the antibiotics prescribed at the time of tissue sample collection and at the time of the final culture report. The antibiotics prescribed were compared with the local prescribing guidelines (Table A1). Each patient's presenting condition was graded retrospectively by the authors based on the information available in the clinical record and using the IDSA classification system 4. Osteomyelitis was classified as a discrete entity.

The tissue samples were collected either intraoperatively in the theatre or by an experienced podiatrist using an aseptic technique after debridement of superficial slough. They were transported promptly to the laboratory and processed according to the local operating procedure guidelines, as supported by the national standard for microbiology investigations 6. Briefly, after the tissue specimen was homogenised, a Gram film was prepared from the fluid and a drop inoculated onto the primary culture plates; the remaining fluid (or a minimum of 1 ml) was inoculated into enrichment broth.

Results

During the study period, 114 samples from 71 patients (76% male) were examined. The mean age of patients was 62·8 years (range 27–88). Of all the samples, 72% were collected through surgical intervention in the operating theatre and 21% from outpatient clinics by the podiatry team, whilst the rest were taken on hospital wards, reflecting the mixed case load in hospital inpatient and outpatient practice. Grading of the ulcers showed 10·5% to be mildly infected (grade 2), 40·3% to be moderately infected (grade 3) and 14·9% to be severely infected (grade 4). Osteomyelitis was present in 27·2% of cases. The remaining 7·0% could not be classified with the available information.

Microscopy and Gram stain

In 57 (50%) of the samples, either no organism or no predominant organism (where a variety of Gram‐positive and Gram‐negative cocci and rods were seen) was reported. Of those where organisms were seen, Gram‐positive cocci were the most common (37·7% of the total samples); Gram‐negative rods were present in 8·8% and Gram‐positive rods in 3·5% of the samples.

Compared with the gold standard of final culture results, sensitivity and specificity of Gram staining were low, at 74·5% and 69·8%, respectively (Table 1). For Gram‐positive cocci, the sensitivity was 54·7%, and the specificity was 77·1%. For Gram‐negative rods, the sensitivity was particularly poor (17·0%), but the specificity was excellent at 98·5%. Overall, only 44·7% (n = 54) of the Gram stain findings were in complete agreement with the culture results.

Table 1.

Sensitivities and specificities of initial microscopy (pus cells and Gram stain) results compared with final bacterial culture results (n = 114)

	Sensitivity (%)	Specificity (%)
Pus cells	64	31
Gram‐positive cocci	55	77
Gram‐positive rods	17	96
Gram‐negative rods	17	99
Overall Gram stain	75	70

Pus cells were reported in only 64% of the samples with positive culture, whereas 69% of the samples in which pus cells were seen did not yield any organisms in culture (Table 1). Thus, the presence of pus cells on microscopy had a positive predictive value of 85·9% and a negative predictive value of just 31·0%.

Culture results

A total of 93 (81·6%) samples yielded a positive culture, with up to four organisms grown in each sample (Figure 1A). At least one Gram‐positive organism was isolated in 59 (50%) samples and at least one Gram‐negative in 47 (39·8%). A total of 24 (20·3%) samples grew both Gram‐positive and ‐negative organisms. In 21 (18·6%) samples, there was no growth.

Figure 1.

Results of bacterial cultures from tissues samples. (A) The range, in number, of organisms grown from each sample. (B) The proportion of cultures found to be polymicrobial compared with the grade of ulcer and (C) the type of organisms grown compared with the grade of ulcer. OM, osteomyelitis.

The organisms most commonly isolated were anaerobes (n = 27) and Staphylococcus aureus (n = 19, 16·7%), followed by Escherichia coli (n = 13, 11·4%) and coagulase‐negative staphylococci (n = 12, 10·5%) (Table A2). A total of 11 S. aureus isolates (7·1% of all isolates) were found to be methicillin‐resistant (MRSA). In three patients, this represented a first MRSA isolation. Vancomycin‐resistant Enterococcus (VRE) was isolated in two patients, neither previously known to be colonised with this multi‐resistant organism. Polymicrobial growth was not associated with any particular grade of infection (Figure 1B), and there was no significant difference between the classes of organisms isolated from the different ulcer grades (p = 0·156) (Figure 1C).

Antibiotic prescribing

Of the 71 patients (representing 106 separate samples), 65 were further analysed for prescribing data; six patients were either lost to follow‐up, or insufficient antibiotic data was obtained. At the time the sample was taken, 72% of patients were already receiving antibiotics, but in only 31·1% was the antibiotic decision compliant with the local prescribing guidelines and previous microbiology isolates (Table A1). The antibiotic regime was changed for 64·1% of patients after the final culture report was received. Following the final culture report, 86·8% of patients were found to be receiving the appropriate antibiotic regimen.

Empirical prescribing was diverse, with 28 different combinations of antibiotic regimes given. The single most common regime was monotherapy coamoxiclav (15·9%), followed by flucloxacillin (12·2%). These are in keeping with the recommended local treatment regimens for mild and moderate DFI (Table A1).

Discussion

Although international guidelines recommend microscopy and Gram staining of soft tissue and bone biopsies taken from the infected diabetic foot 1, 4, in this study, we provide evidence that simple microscopic analysis of tissue samples is generally a poor indicator of culture results, which are the current gold standard for identifying pathogenic organisms in DFI. Neither the Gram stain in assessing for the causative organism nor the presence of pus cells as a marker of infection was overall particularly helpful. However, the high specificity of the Gram stain for Gram‐negative rods (98·5%, section Microscopy and Gram stain) suggests that the Gram stain may be highly useful for ruling in an infection with one of these pathogens. This is clinically significant as this would indicate the necessity of broadening antibiotic therapy beyond Gram‐positive cover should Gram‐negative rods be seen.

Our findings are in contrast with those from a study in Tanzania, reporting an overall 94% sensitivity of the Gram stain compared with culture 7. However, patients in that study had received no prior antibiotic treatment, whereas in our cohort, the majority of samples (72%) were collected from patients who were already undergoing antibiotic therapy (section Antibiotic prescribing). Prior anti‐infective treatment clearly influences the microscopy and culture findings but is nevertheless representative of current clinical practice in the UK, when patients are likely to have already been prescribed antimicrobials, either by primary or secondary care clinicians, before specimens are obtained. Stopping antimicrobials for 10–14 days prior to the tissue biopsy in order to improve microbiological culture yield is often not deemed safe whilst managing an active DFI. Furthermore, although in the previous study, an aseptic surgical technique was used to collect the samples, it is not described if and how the ulcer was cleaned beforehand. In our cohort, 72% of samples were taken within an operating theatre following debridement of the wound, and the remaining samples were collected in the outpatient department by experienced trained podiatrists. Finally, in addition to the routine culture of tissue samples in the study by Abbas et al., our standard laboratory practice is to also use enrichment broth culture, further increasing the final yield.

In our hospital, the microbiology laboratory is situated on site and benefits from a specimen delivery system that minimises the time required for specimens to reach the laboratory. The majority of samples in our study were indeed processed the same day (data not shown), but it is commonly acknowledged that delays in processing microbiological samples influence culture results.

An interesting recent paper by Oates at al. showed that bacterial microcolonies and putative biofilm matrix could be visualised on debridement samples from chronic foot wounds of diabetic patients, which were histologically sectioned and examined by Gram stain 8.

The laboratory diagnosis in DFI is evolving, with the potential for molecular techniques with high sensitivity to alter the way we understand and treat the condition 9. The relatively low overall sensitivity and specificity of microscopy is not altogether surprising as this technique's performance depends on the laboratory operator and the quality of the sample. These are inherent challenges for microscopic techniques. Given this, the role of microscopic analysis of samples in informing treatment appears unclear in a well‐resourced service. However, practical considerations, such as speed and cost, mean these techniques will continue to be important, particularly in low‐resource areas where providing a full range of bacterial cultures may be challenging, and newer techniques will likely prove financially prohibitive. In settings where microscopy is used, operator expertise in preparing and analysing samples will continue to be of vital importance.

The IWGDF strongly recommend that bacterial culture be used to identify the aetiology of a DFI through good‐quality tissue samples 1, 4. The culture of tissue samples has been shown to be superior in identifying pathogens, although the presence of colonisers reduces the specificity of the test 1. Although this is a small retrospective series, our results add to the existing body of data on the bacterial aetiology of diabetic foot ulcers. Of particular note was the isolation of MRSA and VRE in patients who were not previously known to be colonised with these multi‐resistant pathogens (section Culture results). This highlights the importance of undertaking the culture, especially in the current era of increasing antimicrobial resistance, as commonly prescribed empirical treatments would not provide adequate cover for these multi‐resistant pathogens, and it has been shown that as long as antibiotic treatment is adequate, the prognosis in infections with such organisms may not necessarily be any worse 1, 10.

While large studies from Europe and the USA 1, 11 demonstrate that Gram‐positive cocci are the predominant aetiological organisms, in our study, anaerobic organisms were the most common isolated pathogens (section Culture results). This may be because of improved laboratory isolation techniques, such as using enrichment culture and extending the period of incubation. We further confirm the polymicrobial nature of the DFI, and we therefore recommend that all microorganisms isolated be identified at the species level rather than dismissing polymicrobial cultures as ‘mixed growth’ 11. However, unlike previous research in this area 12, 13, we did not find an association between the grades of ulcer and any specific changes in the bacteriology of the samples, although this may be because of the low number of samples included in this study (section Culture results). A direct comparison with earlier studies, however, is difficult because of the variety of grading systems used to classify ulcers in these reports. Given the more widespread adoption of the IWGDF and IDSA classification schemes in many settings, such a comparative analysis may become easier in the future 5, 14, 15.

There is now an increasing amount of data on the service provision and outcomes of patients with diabetes in the UK because of initiatives such as the National Diabetes Inpatient Audit 16. The anticipated National Diabetes Foot Care Audit may provide further insight into UK practice 17. However, there are still gaps in our knowledge, particularly in how clinicians prescribe antibiotics. Our data shows empirical prescribing to be consistent with local guidelines in only a third (31·1%) of the patients in our hospital. However, we demonstrate that antibiotic treatment was adjusted when culture results became available, with a large improvement (86·8%) at the targeted prescribing stage (section Antibiotic prescribing). This is reassuring and not entirely surprising, but to our knowledge, there has been only one other report that documents this behaviour; a recent study from the USA showed similar results, with a higher correct empirical prescribing of 59·7% but lower targeted antibiotic usage in 73·7% of patients 18. It follows, therefore, that clinicians are using microbiological data and advice to alter their prescribing patterns. We suspect that the main reasons for suboptimal empirical therapy are not only the fact that staff are unfamiliar with local guidelines but also a lack of general awareness of the IDSA classification system and IWGDF recommendations. The IDSA classification has been validated and it has a significant prognostic value, with higher grades of infection being associated with increased rates of major amputation as well as increased length of hospital stay 19, 20, 21.

Our study has several limitations. The DFI severity was graded retrospectively and was therefore heavily reliant on the documented information. However, we were able to harness the benefit of the hospital electronic patient system, which provides the contemporaneous results of patient physiological observations and prescribing data, allowing comprehensive data analysis.

Our hospital is a regional referral centre for vascular disease, which likely reflects the caseload mix and the likely presence of more complex infections than in other patient populations. The high proportion of samples collected in the operating theatre illustrates this.

Histopathological analysis of bone samples in cases of suspected osteomyelitis is seen as the optimal test for diagnosis in conjunction with microbiological data 22. In our centre, as in others, this examination is not routinely undertaken, and a recent review noted that bone biopsies are rarely carried out 1, 23.

Conclusion

In conclusion, microscopy was generally a poor predictor of tissue culture results. However, as it is a quick and inexpensive investigation, it maintains a role within the laboratory repertoire of investigations for DFI. Furthermore, here, we have identified a narrow but clinically important potential role for it in the identification of Gram‐negative rods that could allow early broadening of antibiotic treatment. We have also shown that despite initial poor compliance of empirical antibiotic treatment regimens, prescribing was adjusted in light of culture results, suggesting that these were important for clinicians in our hospital.

Table A1.

Local antibiotic prescribing guidelines for diabetic foot infection

Infection severity IDSA grade (IWGDF classification)	First choice	Second choice	MRSA positive
Mild (grade 2)	Flucloxacillin 1 g QDS PO	Clindamycin 450 mg QDS PO	Vancomycin 1 g BD IV or doxycycline 200 mg Stat, 100 mg BD PO
Moderate (grade 3)	Coamoxiclav 625 mg TDS PO or 1·2 g TDS IV	Clindamycin 300–450 mg QDS PO or 300–600 mg QDS IV and Ciprofloxacin 500 mg BD PO or 400 mg BD IV	As second choice or vancomycin 1 g BD IV and rifampicin 600 mg BD PO and ciprofloxacin 500 mg BD PO or 400 mg BD IV or replace rifampicin with sodium fusidate 500 mg TDS PO
Severe (grade 4)	Meropenem 1 g TDS IV	Clindamycin 600 mg QDS IV and ciprofloxacin 400 mg BD IV	Meropenem 1 g TDS IV and vancomycin 1 g BD IV and rifampicin 600 mg BD PO
Osteomyelitis (OM)	Flucloxacillin 2 g QDS IV and ciprofloxacin 750 mg BD PO and metronidazole 400 mg TDS PO	Clindamycin 450 mg QDS PO or 600 mg QDS IV and ciprofloxacin 750 mg BD PO	Vancomycin 1 g BD IV and ciprofloxacin 750 mg BD PO and metronidazole 400 mg TDS PO

IDSA, Infectious Disease Society of America; IWGDF, International Working Group for Diabetic Foot; BD, twice daily; TDS, thrice daily; QDS, four times a day; IV, intravenous; PO, per os (orally).

Table A2.

Organisms isolated

Organism	Number of tissue samples from which organism was isolated
Anaerobes	25
Staphylococcus aureus	19
Methicillin‐resistant S. aureus	11
Methicillin‐sensitive S. aureus	8
Escherichia coli	13
Coagulase‐negative Staphylococcus	12
Enterococcus sp.	10
Vancomycin‐sensitive Enterococcus	9
Vancomycin‐resistant Enterococcus	2
Beta‐haemolytic Streptococcus group B	8
Proteus mirabilis	8
Corynebacterium sp.	7
Enterobacter cloacae	6
Serratia marcescens	6
Enterococccus faecalis	5
Pseudomonas aeruginosa	5
Morganella morganii	4
Klebsiella pneumoniae	3
Proteus sp.	3
Beta‐haemolytic Streptococcus group G	3
Klebsiella oxytoca	2
Peptostreptococcus	2
Proteus vulgaris	2
Alcaligenes faecalis	1
Bacteroides	1
Candida sp.	1
Citrobacter freundii	1
Citrobacter koseri	1
Clostridium sp.	1
Pediococcus sp.	1
Providencia rettgeri	1
Streptococcus agalactiae	1
Staphylococcus epidermidis	1
Streptococcus milleri	1
Streptococcus mitis	1
Unidentified Gram‐negative rod	1