Concordance in diabetic foot ulceration: a cross-sectional study of agreement between wound swabbing and tissue sampling in infected ulcers

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 09/75/01. The contractual start date was in April 2011. The draft report began editorial review in February 2015 and was accepted for publication in July 2015. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Professor E Andrea Nelson was a member of the Health Technology Assessment Commissioning Board.

Copyright statement

© Queen’s Printer and Controller of HMSO 2016. This work was produced by Nelson et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Diabetes: prevalence and complications

Worldwide prevalence of diabetes mellitus was estimated at around 2.8% in 20001 and this is predicted to increase to 7.7% (affecting 439 million adults) by 2030,2 largely as a result of the obesity epidemic. 3,4 In the USA, the prevalence of diabetes was 8.3% in 2012, which is a sharp increase (more than doubling) compared with the prevalence in 1990, which was 3.5%. 5 Estimates from the USA predict that Americans born in 2000 will have a one in three lifetime risk of developing type 2 diabetes. 6 In the UK, the prevalence of diabetes is 6.0% (3.2 million people),7 and Diabetes UK estimates that there are around 630,000 people with diabetes who have not yet been diagnosed. 8 Treatment of diabetes in the UK cost approximately £23.7B in 2010/11, accounting for approximately 10% of the total health resource expenditure.

Both type 1 and type 2 diabetes can lead to serious health problems. 9 Complications of diabetes, especially in patients with poorly controlled blood sugar levels, include damage to the eyes, kidneys, nerves and arteries. In the feet, diabetes-related peripheral neuropathy leads to changes to foot architecture (hence increasing pressure on plantar surfaces, including those unaccustomed to load-bearing),10,11 reduced sweating (hence dry, cracking skin), poor sensation (hence susceptibility to trauma) and accelerated atherosclerotic disease, which leads to reduced circulation, with consequent problems with healing wounds and fighting infection. These peripheral neuropathic and vascular changes, either alone or in combination, predispose the foot to ulceration and its sequalae. 12,13

Diabetic ulcer infection: epidemiology and aetiology

It is estimated that the proportion of people with diabetes in the UK who have ever had a foot ulcer is around 6%,14 and that lifetime risk of ulceration is 15–25%. 15,16 Diabetic foot ulcers (DFUs) can take many weeks or often months to heal and have a negative impact on patients’ functional ability and quality of life. Foot ulceration in diabetic individuals also has a wider societal impact, such as reduced work productivity, high health-care costs and financial loss. 17–21 An open wound related to foot ulceration, combined with various immunological perturbations caused by diabetes, frequently results in infection. Prospective studies have found that about half of recent-onset DFUs are clinically infected at presentation. 22 Diabetic foot infection is thought to be the most common cause of diabetes-related hospital admissions and precedes approximately 80% of non-traumatic lower limb amputations. 15,23–27

Foot infections in people with diabetes can be hard to manage because of the associated impaired arterial supply to the legs, as well as impaired function of the immune system (especially those related to defects in polymorphonuclear leucocytes). This leads to an increased risk of progression of infection with contiguous spread to deeper tissues (including bone), and proximal extension up the foot and leg, as well as systemic spread into the blood stream. Therefore, many diabetic foot infections require some level of lower extremity amputation as a limb-sacrificing, but potentially life-saving, measure. 14,28,29

The incidence of lower extremity amputations is 10–30% higher in people with diabetes than in the general population,28,29 and about 85% of these amputations are preceded by a foot ulcer infection. 23,24,26,30–33 Limb amputation is associated with major consequences, as it dramatically reduces health-related quality of life, is expensive for both the patient and the health-care system and is associated with a 5-year mortality of over 50%. 15 To reduce the risk of foot ulceration, accelerate the healing of open ulcers and identify and treat infection promptly, many health-care systems have deployed multidisciplinary foot teams to co-ordinate foot care. The prevention of foot ulceration and amputation involves optimising glycaemic control and foot care. This may include supplying pressure-relieving shoes or insoles, undertaking surgical interventions promptly and optimally treating any infection. Providing this care involves input from the specialties of general practice, diabetology, nursing, dietetics, podiatry, orthotics, vascular surgery and infectious diseases/clinical microbiology.

Wound infection: definition, identification and characterisation

All chronic wounds, including DFUs, have bacteria on their surface that may originate from the surrounding normal skin flora, as well as opportunistic bacteria, such as gut flora. Therefore, the presence of bacteria in a wound does not indicate the presence of infection. When the host tissues show no inflammatory response or incur no damage associated with the bacterial growth, then the wound is described as ‘colonised,’ rather than infected. At this stage, there is typically thought to be a ‘balance’ between the growth of the several species of bacteria and no single organism usually dominates. When a critical density or high virulence of colonising organisms causes damage to host tissues, the wound is deemed to be ‘infected’. Therefore, infection of chronic wounds is usually a clinical diagnosis based on signs and symptoms of host tissue inflammation, such as pyrexia, purulent secretions, pain or tenderness, erythema, warmth and induration. 34–38 Although some investigators and clinicians also describe a quantitative diagnostic criterion for the presence of infection (e.g. a bacterial load of > 10⁵ colony-forming units per gram of tissue), there is no agreement on this. 39 In chronic wounds, a single cut-off point for bacteria has been found to be insufficient for defining infection; other factors, such as the number or virulence of the bacterial groups and the presence of biofilm, are also important. 40

When a wound infection is diagnosed, the therapeutic approach depends on the whole clinical situation. Because in infected diabetic foot wounds the consequences of delayed antibiotic treatment can be profound, empiric antibiotic therapy should usually be initiated immediately. The antimicrobial regimen is usually selected in accordance with departmental protocols that are based on the probable causative organisms and their susceptibility patterns. Concurrently, samples for microbiological analysis are taken to identify the infecting organisms within the wound and their susceptibility to a range of antimicrobials. The resulting microbiological information guides subsequent modifications of the empiric antibiotic therapy required should the infection not improve and resistant organisms be isolated. 34–38,41,42 The culture and sensitivity results also allow a change from broad- to narrow-spectrum antibiotic agents, thus following the principles of antibiotic stewardship. 34–38,41

The microbiological analysis of specimens from the ulcer is useful only if the specimen is properly collected and processed and reported accurately and promptly. The aim is to acquire a wound sample that identifies all pathogens while avoiding colonising flora. First, the ulcer must be cleaned, which may involve debridement to remove necrotic material or callus and undermining tissues. Second, a specimen is taken from the site of infection, using one of a number of specimen-acquisition techniques, such as wound swabbing, fluid sampling using a fine-needle aspiration, or tissue sampling (by biopsy or curettage). 36,37,42 Taking a tissue sample either uses a tool to extract a ‘punch biopsy’ or scrapes the base of a wound with a sharp-edged dermal curette or scalpel blade to obtain ulcer tissue from the debrided ulcer bed. 43

It is important that the culture of the sample obtained reflects an accurate profile of the bacterial environment in the ulcer. Either failing to identify a true pathogen or identifying a coloniser as a pathogen can lead to inappropriate treatment of an infected wound. Therefore, it is important that health-care staff use a technique that will give a specimen that provides an accurate account of the bacteria present, including their number and sensitivity to antibiotics. Most published guidelines recommend obtaining a tissue specimen rather than a swab, in order to increase the likelihood of accurately reflecting the organisms associated with clinical infection at initial presentation. 34,36,37,42

In clinical practice, however, samples from wounds are often taken with a cotton swab. 44–47 The advantages of a wound swab include the almost universal availability of the equipment, the relative ease of the technique, the low cost of the swab and the fact that little training is needed to perform this correctly, which means that it can be done by non-clinician staff. 45 Furthermore, there is little risk of harm using a swab to collect a tissue sample. The disadvantages of a swab include the concerns that it may not collect those bacteria responsible for the infection deep within the tissues (e.g. as happens if an appropriate technique is used), that it will collect the colonising bacteria on the wound surface, or that it will fail to provide an environment conducive to growth of obligate anaerobes and other fastidious organisms (i.e. those that may be present in the wound but die in a swab device that does not provide an adequate medium for their survival). To counter these problems, advocates of wound swabbing have specified how to prepare the ulcer bed (i.e. removing dead tissue that may contain non-pathogenic bacterial groups) and how to obtain a sample from deep in the ulcer (by pressing to collect fluid from deep in the subcutaneous tissues, as described by Levine et al. 48 in 1976), as well as the optimal storage and transport procedures (use of charcoal swab, transport medium and swift delivery to the laboratory to maintain the viability of fastidious organisms). 48

In contrast, the reported advantage of tissue sampling is that the specimen is likely to contain the pathogens responsible for tissue destruction and infection. However, tissue-sampling techniques require disruption or cutting of the ulcer bed to obtain a specimen and this may lead to bleeding or pain (although most DFUs are complicated by neuropathy, which reduces the ability to perceive pain). Some clinical staff may need additional training to be able to take these samples safely and they also require some basic equipment: sharp sterile blades (scalpel), dermal curettes or a biopsy cutter. Using appropriate storage and transport procedures (transport medium and swift delivery to laboratory to maintain the viability of fastidious organisms) is still required.

Diabetic foot ulcer guideline recommendations for infection (diagnosis/identification and characterisation and treatment)

Several guidelines and consensus documents aimed at improving the care for people with DFUs have been published over the past decade. 9,15,34–37,41,42,54,55 In this report, we have focused on three guidelines: (1) the UK National Institute for Health and Care Excellence (NICE) guidance on inpatient management of diabetic foot problems;37 (2) the Infectious Diseases Society of America (IDSA) guidelines for the diagnosis and treatment of diabetic foot infections;1,2,36 and (3) the International Working Group on the Diabetic Foot (IWGDF) guidelines on the management and prevention of the diabetic foot. 42,56 IDSA guidelines were first published in 200434,35 and are widely used. The IWGDF guidelines were published in 200842 and the latest NICE guidance in 2011. 37 The IDSA guidelines have recently been updated and provide details on the strength of the recommendations and the quality of the supporting evidence,36 making them the most current and comprehensive guidelines for the diagnosis and management of DFUs.

National Institute for Health and Care Excellence guidance37 recommends that clinicians should evaluate a diabetic patient presenting with a foot wound at three levels: the patient as a whole, the affected foot or limb and the infected wound. For infected wounds, an appropriately obtained specimen for culture is recommended prior to starting empiric antibiotic therapy, if possible. NICE guidance37 recommends sending a specimen for culture that is from deep tissue, obtained by biopsy or curettage and after the wound has been cleansed and debrided. The guidance advises against taking swab specimens, especially of inadequately debrided wounds, as they are likely to provide less accurate results. The IWGDF guidelines have the same message about obtaining the specimen but also mention the value of obtaining a Gram-stained smear of the wound in addition to culture. 42 For infected wounds, the IDSA guidelines34–36 recommend sending a specimen for culture that is from deep tissue, obtained by biopsy or curettage after the wound has been cleansed and debrided. The guidelines also advise against obtaining for culture by swabbing the wound or wound drainage. 36 In summary, all the clinical guidelines34–37,42 agree on their preference of tissue sample (obtained by biopsy, curettage or aspiration) to wound swab specimens.

The need for research

Although clinicians commonly use swab samples to provide information on the bacteria in a clinically infected wound, the current major guidelines all recommend tissue specimens over swab samples. 34–38 This is mainly because swabs can be contaminated with colonising flora, can miss deep pathogens and may be less likely to grow anaerobic and some fastidious aerobic organisms. However, the strength of this recommendation was specified only in IDSA guidelines,36 where it was ‘moderate’ (i.e. further definitive research is likely to have an important impact on future recommendations). 57

Three primary studies of culture techniques informing the guidelines were those conducted by Pellizzer et al. ,58 Slater et al. 59 and Bill et al. 60 Pellizzer et al. 58 assessed the reliability of results of ulcer swabbing versus deep tissue biopsy in 29 diabetic patients with a limb-threatening foot infection, who were neither recently treated with antibiotics nor hospitalised. This selected population does not reflect many of the patients with foot infections seen in outpatient clinics, who have often had recent antibiotic therapy. The study did not report on the agreement between swab and tissue samples, but, rather, simply on the number of bacterial colonies in each. Their conclusion that tissue samples are better than swab samples was based on a comparison of the numbers of isolates in only 21 participants remaining in the study at 30 days. Their finding may be due to chance as they performed 20 comparisons without adjustment for multiple testing. Furthermore, a method that identifies more colonies may be collecting more colonising bacteria and, therefore, is not necessarily ‘better’. The unpaired analysis presented means that we cannot readily compare the two techniques using appropriate statistical methods.

Slater et al. 59 aimed to evaluate the accuracy of swab compared with deep tissue (obtained via needle aspiration) cultures in diabetic wounds of varying depth and severity. Their study, however, included only 30 people with ulcers (in a sample of 60, in which the other patients had deep abscesses, etc.) and it is not clear if the results were heterogeneous across types of wounds or apply to tissue samples collected using scalpel or curette. In 62% of the samples, there was a similar profile of organisms isolated from the swab and the deep tissue sample, whereas in 20% of samples the swab identified more organisms and in 18% the deeper tissue sample picked up more organisms. These data were not stratified by the presence or absence of an ulcer or by ulcer type (i.e. neuropathic or ischaemic). This study identified that there can be two forms of disagreement between swabbing and sampling, with swabs identifying more organisms or tissue samples identifying more organisms; hence, they did not consider either technique to be a gold standard.

In a 2006 systematic review of the diagnosis and management of infection in DFUs,61 only one study that evaluated sample acquisition and reported agreement in sufficient detail to allow appropriate analysis was identified. This study by Bill et al. 60 included 18 patients with a pressure ulcer, 10 with a DFU, 5 with a venous leg ulcer, and 5 with an arterial ulcer. In this study, quantitative analysis of bacterial growth from a punch biopsy taken from the centre of the wound was compared with that of a wound swab. Using a definition of infection of a bacterial load of > 10⁶ bacteria per gram of tissue in the punch biopsy, the authors reported a sensitivity for wound swabbing of 79%, meaning that the swab failed to detect approximately one in five wound infections as defined by punch biopsy. The derived likelihood ratios suggested that the wound swab was not a useful method of identifying infection in chronic wounds. Interpretation of this study’s findings is impeded by its small size and heterogeneity in the ulcer population. We cannot be sure that these data are directly transferable to the population of interest here, namely people with a DFU and a clinically diagnosed ulcer infection (there is no reason to sample uninfected ulcers and inclusion of people with uninfected foot ulcers may reduce the external validity of the study). In addition, there were potential sources of bias, such as no description of blind test verification and lack of clarity over whether or not the same clinical data were available when test results were interpreted as would be available when the test is used in practice.

Two studies62,63 have been published since the IDSA, NICE and IWGDF guidelines. Mutluoglu et al. 62 assessed the reliability of cultures of superficial swabs by comparing them with cultures of concomitantly obtained deep tissue specimens in patients with DFUs. They retrospectively reviewed the notes from 54 patients from whom there were 89 pairs of samples, one a superficial swab and the other deep tissue. The results showed a 73% concordance between swab cultures and deep tissue biopsies, which dropped to 69.2% when sterile pairs of cultures were excluded. Compared with deep tissue specimens, in 11.2% of cases swabs detected additional species, in 9.0% of cases swabs detected fewer species and in 6.7% the two techniques identified totally different organisms. The study concluded that superficial swab cultures are not sufficiently accurate to identify the causative organisms in patients with an infected DFU. They described three forms of disagreement: swabs identified more organisms, tissue samples identified more organisms and the techniques found different organisms.

Demetriou et al. 63 assessed the diagnostic performance of swabs versus tissue cultures in 50 consecutive diabetic patients with a foot ulcer, 28 of which were neuropathic and 22 of which were neuroischaemic. The authors stated that 36 (72%) wounds were infected, based on ‘the presence of at least 2 of the following criteria: local swelling or induration, erythema greater than 0.5 cm in any direction around the ulcer, local tenderness or pain, local increase of temperature, and purulent discharge’. Overall, the results showed that swabs reported significantly more isolates than tissue cultures, this difference being more evident in neuropathic than in neuroischaemic ulcers. They defined the tissue sample as the ‘gold standard’ for the diagnosis of infection, and reported swab culture sensitivity of 100% and negative predictive value of 100%, but the specificity was only 14.3% in neuropathic and 18.2% in neuroischaemic ulcers. They concluded that swabs are useful only to rule out infection. Given that guidelines do not recommend sampling/swabbing uninfected ulcers, the inclusion of 14 people in this study with uninfected ulcers reduces its external validity.

In summary, we concluded that the existing evidence regarding the results of cultures of specimens obtained by swabbing versus tissue sampling was derived from small, heterogeneous and, often, methodologically poor studies. Thus, there is a lack of robust evidence on the most appropriate method to use in routine clinical practice.

The question addressed in this study was not how to diagnose infection in a DFU, but rather what was the best way to collect a sample to characterise the bacterial flora. We therefore set out to describe the patterns of agreement and disagreement between swab and tissue samples. To help advise clinicians on the best technique to identify pathogens and to avoid colonising organisms in DFUs, we conducted a series of studies. The first was the ‘main study’, followed by three substudies:

• main study: cross-sectional study to determine the patterns of agreement between culture results of contemporaneously collected swab versus tissue samples

• substudy 1: independent ‘virtual’ clinical review of the appropriateness of empirical antimicrobial therapy based on the results of swabs compared with tissue samples to describe the potential clinical relevance of any differences in sampling results from swabs and tissue

• substudy 2: a pilot comparative study of results of standard plating and culture techniques versus the molecular technique of PCR

• substudy 3: a study of the prognosis of diabetic foot infection.

The main study was the first large, cross-sectional, multicentre study to examine agreement and disagreement of culture results between swab and tissue sampling techniques taken at the same time in a large group of patients with a clinically infected DFU. 64 Each of the studies is described in detail in Chapters 2–5.

Study aims

The primary aim of the main study (patterns of agreement between swab sampling and tissue sampling) was to evaluate concordance between culture results from wound swabs and tissue samples from the same patient (see Chapter 2).

The aim of the clinical panel review study was to evaluate whether or not any differences in bacterial profiles from specimens obtained from swabs and tissue samples are clinically relevant. This was done by ascertaining from a panel of clinicians whether or not the reports from a swab or tissue sample would have resulted in a change in clinical management (see Chapter 3).

The aim of the pilot comparative study of standard plating and culture techniques versus PCR was to assess the concordance between results from specimens taken by conventional culture techniques and by molecular techniques (see Chapter 4).

The aim of the prognosis of foot infection study was to determine the outcome of patients with an infected DFU at 12 months post registration and to explore prognostic factors that may be related to time to wound healing (see Chapter 5).

Patient and public involvement

In this study, patient and public involvement was achieved by using our links with diabetes organisations at the national (Diabetes UK), regional (North West Diabetes Local Research Network) and local (School of Healthcare Service User Group) levels.

As this work was commissioned by the NHS Health Technology Assessment (HTA) programme, then there had been patient and public involvement engagement at the prioritisation stage, and this informed the commissioners as regards the importance and relevance of the clinical question.

During the study we were fortunate to recruit a patient representative, Mrs Christine Thomas, as a member of the Study Steering Committee (SSC). She played a key part in the SSC meetings and advised the study team at different stages, including at the writing of patient and public-facing information. She also had an important role in shaping all aspects of the communications with patients as regards consent, particularly when moving to verbal consent. Furthermore, Mrs Thomas advised the study team about the dissemination of the initial results to participants at the end of the study and reviewed draft communications.

Introduction

For infected DFUs, the accurate identification of pathogens, rather than colonising bacteria, is a prerequisite for selecting targeted antibiotic therapy to ensure optimal patient outcomes and avoid the acquisition of antibiotic resistance. Currently available evidence from the main diabetic foot infection guidelines (NICE,37 IDSA1,2,36 and the IWGDF42,56) and other studies62,63 is not sufficiently robust to advise clinicians on the best technique to identify pathogens in DFUs.

Methods

Study design

A multicentre, cross-sectional study involving 400 patients with a DFU with suspected infection requiring antibiotic therapy was conducted (Figure 1). Consenting patients had both a swab and tissue sample taken from their suspected infected DFU for conventional plating and culture.

FIGURE 1.

Study flow diagram. Adapted from Nelson EA, Backhouse MR, Bhogal MS, Wright-Hughes A, Lipsky BA, Nixon J, et al. Concordance in diabetic foot ulcer infection. BMJ Open 2013;3:e002370. Used under Creative Commons Attribution-Non Commercial 2.0 licence. Co-I, co-ordinator; PI, principal investigator.

Results

Sample size

In total, 680 patients were screened for recruitment into CODIFI and 401 patients were enrolled between November 2011 and May 2013. One patient was excluded as the written informed consent was lost. Of 27 centres, 25 recruited patients into the study; Figure 2 shows the number of patients recruited and screened per centre, and Figure 3 shows the overall, monthly and cumulative recruitment of patients to the study together with our original target. See Appendix 1 for full centre names.

FIGURE 2.

Screening and recruitment by centre. SJUH, St James’s University Hospital.

FIGURE 3.

Monthly and cumulative recruitment.

Study conduct

Figure 4 presents a flow diagram depicting the study conduct and analysis population.

FIGURE 4.

Patient flow diagram.

Baseline characteristics

Tables 5–13 summarise the baseline characteristics, including patient demographics, information about diabetes, clinical details, ulcer characteristics, PEDIS classification, clinical signs and symptoms, and antibiotic regimens immediately pre and post sampling. Because the evaluable population was very similar to the full analysis set with respect to baseline characteristics, characteristics of the full sample only are detailed below.

TABLE 5 - Baseline patient demographics and type of recruiting site

SD, standard deviation.

Patient demographics	Full analysis set (N = 400)	Evaluable population (N = 395)
Age (years)
Mean (SD)	63.1 (13.3)	63.1 (13.4)
Median (range)	63.0 (26–99)	63.0 (26–99)
Missing	0	0
Sex, n (%)
Male	316 (79.0)	311 (78.7)
Female	84 (21.0)	84 (21.3)
Ethnicity, n (%)
White	377 (94.3)	372 (94.2)
Other mixed background	1 (0.3)	1 (0.3)
Asian: Indian	3 (0.8)	3 (0.8)
Asian: Pakistani	11 (2.8)	11 (2.8)
Other Asian background	2 (0.5)	2 (0.5)
Black: Caribbean	3 (0.8)	3 (0.8)
Black: African	1 (0.3)	1 (0.3)
Other ethnic group	2 (0.5)	2 (0.5)
Site of recruitment, n (%)
Hospital ward	53 (13.3)	53 (13.4)
Outpatient clinic	319 (79.8)	314 (79.5)
Community clinic	28 (7.0)	28 (7.1)

TABLE 6 - Baseline diabetes details

HbA_1c, glycated haemoglobin; SD, standard deviation.

Diabetes details	Full analysis set (N = 400)	Evaluable population (N = 395)
Diabetes type, n (%)
Type 1	58 (14.5)	58 (14.7)
Type 2	342 (85.5)	337 (85.3)
Duration of diabetes (years)
Mean (SD)	16.8 (11.0)	16.9 (11.0)
Median (range)	15.0 (0.04–57)	15.0 (0.04–57)
Missing	3	3
HbA1C (%)
Mean (SD)	8.72 (2.29)	8.71 (2.29)
Median (range)	8.10 (4.6–17.2)	8.10 (4.6–17.2)
Missing	6	6
Current diabetes treatment, n (%)
Yes	385 (96.3)	381 (96.5)
No	15 (3.8)	14 (3.5)
Diabetes treatment details, n (%)
Oral hypoglycaemic agent	107 (27.8)	106 (27.8)
Insulin	168 (43.6)	166 (43.6)
Both oral hypoglycaemic agent and insulin	109 (28.3)	108 (28.3)
Oral hypoglycaemic agent and exenatide	1 (0.3)	1 (0.3)

TABLE 7 - Ulcer characteristics

SD, standard deviation.

Ulcer characteristics	Full analysis set (N = 400)	Evaluable population (N = 395)
Location of ulcer(s), n (%)
Ulcers on both right and left foot	60 (15.0)	59 (14.9)
Ulcer(s) on right foot only	173 (43.3)	169 (42.8)
Ulcer(s) on left foot only	167 (41.8)	167 (42.3)
Total number of ulcers, n (%)
1	268 (67.0)	264 (66.8)
2	78 (19.5)	78 (19.7)
3	43 (10.8)	43 (10.9)
4	6 (1.5)	6 (1.5)
5	1 (0.3)	1 (0.3)
6	3 (0.8)	2 (0.5)
7	1 (0.3)	1 (0.3)
Mean (SD)	1.5 (0.9)	1.5 (0.9)
Median (range)	1.0 (1–7)	1.0 (1–7)
Missing	0	0

TABLE 8 - Index ulcer characteristics

SD, standard deviation.

The other locations of index ulcer are the dorsum and digital for two patients and the apex interdigital space plantar surface, lateral surface, lateral outer aspect of foot, left lateral malleolus, medial surface and medial malleolus for the remaining patients.

Index ulcer characteristics	Full analysis set (N = 400)	Evaluable population (N = 395)
Foot containing index ulcer, n (%)
Right foot	205 (51.3)	201 (50.9)
Left foot	195 (48.8)	194 (49.1)
Index ulcer location, n (%)a
Apex	47 (11.8)	45 (11.4)
Interdigital	25 (6.3)	25 (6.3)
Plantar	172 (43.0)	170 (43.0)
Dorsum	56 (14.0)	56 (14.2)
Digital	90 (22.5)	89 (22.5)
Other	8 (2.0)	8 (2.0)
Missing	2 (0.5)	2 (0.5)
Duration of index ulcer (months)
Mean (SD)	5.58 (12.28)	5.52 (12.17)
Median (range)	1.84 (0.1–144.0)	1.84 (0.1–144.0)
Missing	4	4
First or recurrent index ulcer, n (%)
Incident	288 (72.0)	283 (71.6)
Recurrent	110 (27.5)	110 (27.8)
Missing	2 (0.5)	2 (0.5)
Aetiology of index ulcer, n (%)
Ischaemic	14 (3.5)	14 (3.5)
Neuropathic	202 (50.5)	199 (50.4)
Both ischaemic and neuropathic	182 (45.5)	180 (45.6)
Missing	2 (0.5)	2 (0.5)
Antimicrobial dressing on the ulcer, n (%)
Yes	241 (60.3)	238 (60.3)
No	154 (38.5)	152 (38.5)
Missing	5 (1.3)	5 (1.3)

TABLE 9 - Perfusion, Extent/Size, Depth/Tissue loss, Infection, Sensation classification and ulcer debridement

CLI, critical limb ischaemia; PAD, peripheral arterial disease; SD, standard deviation.

Calculated using Kundin:81 length × width × 0.785.

PEDIS classification and ulcer debridement	Full analysis set (N = 400)	Evaluable population (N = 395)
Perfusion, n (%)
Grade 1: no symptoms/signs of PAD	200 (50.0)	197 (49.9)
Grade 2: symptoms or signs of PAD, but no CLI	192 (48.0)	190 (48.1)
Grade 3: CLI	8 (2.0)	8 (2.0)
aExtent/size: estimated index ulcer area, cm2
Mean (SD)	6.76 (15.16)	6.60 (14.85)
Median (range)	1.77 (0.01–138.2)	1.77 (0.01–138.2)
Missing	3	3
Depth/tissue loss, n (%)
Grade 1: superficial full-thickness ulcer	131 (32.8)	130 (32.9)
Grade 2: ulcer penetrating below dermis to skin structures	134 (33.5)	132 (33.4)
Grade 3: all subsequent layers of foot, including bone/joint	135 (33.8)	133 (33.7)
Infection, n (%)
Grade 1: no symptoms/signs of inflammation	2 (0.5)	2 (0.5)
Grade 2: inflammation of skin/subcutaneous tissue only	149 (37.3)	148 (37.5)
Grade 3: extensive erythema deeper than skin/subcutaneous tissue	237 (59.3)	234 (59.2)
Grade 4: systemic inflammatory response syndrome	12 (3.0)	11 (2.8)
Sensation, n (%)
Grade 1: no loss of protective sensation	27 (6.8)	27 (6.8)
Grade 2: loss of protective sensation	373 (93.3)	368 (93.2)
Ulcer debridement undertaken, n (%)
Yes	351 (87.8)	347 (87.8)
No	49 (12.3)	48 (12.2)

TABLE 10 - Clinical signs and symptoms classification

Missing data were present for recent increase in pain, recent increase in wound size and breakdown of epithelium for one patient for each and for recent decrease in pain for three patients.

Presence of clinical signs and symptoms	Full analysis set (N = 400), n (%)	Evaluable population (N = 395), n (%)
Wound odour	127 (31.8)	126 (31.9)
Pocketing in wound	170 (42.5)	168 (42.5)
Discoloured granulation tissue	225 (56.3)	220 (55.7)
Friable granulation tissue	204 (51.0)	202 (51.1)
Recent increase in paina	125 (31.3)	123 (31.1)
Recent decrease in paina	9 (2.3)	9 (2.3)
Recent increase in wound sizea	246 (61.5)	241 (61.0)
Breakdown of epitheliuma	126 (31.5)	124 (31.4)

TABLE 11 - Wagner grade

Wagner grade	Full analysis set (N = 400), n (%)	Evaluable population (N = 395), n (%)
Grade 1: superficial diabetic ulcer (partial or full thickness)	136 (34.0)	135 (34.2)
Grade 2: ulcer extension ligament, tendon, joint capsule or deep fascia without abscess or osteomyelitis	134 (33.5)	132 (33.4)
Grade 3: deep ulcer with abscess, osteomyelitis or joint sepsis	122 (30.5)	120 (30.4)
Grade 4: gangrene localised to portion of forefoot or heel	7 (1.8)	7 (1.8)
Grade 5: extensive gangrenous involvement of the entire foot	1 (0.3)	1 (0.3)

TABLE 12 - Antibiotic therapy: pre and post sampling

SD, standard deviation.

Antibiotic therapy	Full analysis set (N = 400)	Evaluable population (N = 395)
Patient on a pre-sampling antibiotic therapy regimen, n (%)
Yes	187 (46.8)	186 (47.1)
None prescribed	194 (48.5)	190 (48.1)
Missing	19 (4.8)	19 (4.8)
Days spent on pre-sampling antibiotic therapy
Mean (SD)	14.6 (21.9)	14.7 (21.9)
Median (range)	7.0 (1–145)	7.0 (1–145)
Missing	1	1
Change to antibiotic therapy: immediately post sampling, n (%)
Yes	248 (62.0)	244 (61.8)
No	133 (33.3)	132 (33.4)
Missing	19 (4.8)	19 (4.8)

TABLE 13 - Summary of patients’ baseline antibiotic regimen (pre sampling) and proposed new antibiotic regimen (immediately post sampling)

Antibiotic regimen	Full analysis set (N = 400), n (%)	Evaluable population (N = 395), n (%)
No pre-sampling antibiotic but initiation immediately post sampling	168 (42.0)	164 (41.5)
No pre-sampling antibiotic and no initiation immediately post sampling	26 (6.5)	26 (6.6)
On a pre-sampling antibiotic with or without a change immediately post sampling	187 (46.8)	186 (47.1)
Unknown whether or not there was a pre-sampling antibiotic but initiation/change immediately post sampling	19 (4.8)	19 (4.8)

Tables 5 and 6 summarise patient demographics and diabetes details, respectively. The median age of patients was 63 years (range 26–99 years); 79% of patients were male; and the majority of patients (94.3%) were of white ethnic origin. Recruitment of patients was from outpatient clinics for 79.8% of patients, hospital wards for 13.3% and community clinics for 7%. The median duration of diabetes in enrolled patients was 15 years (range 2 weeks–57 years); 14.5% and 85.5% of patients had type 1 or type 2 diabetes, respectively; and the vast majority of patients (96.3%) were receiving treatment for their diabetes.

Tables 7 and 8 summarise patients’ ulcer characteristics. The total number of DFUs ranged from one to seven per patient, with one ulcer observed for 67.0% of patients, two ulcers for 19.5%, and three or more ulcers for 13.6% of patients. The anatomic site of the index ulcer, from which both the swab and tissue samples were obtained, was most commonly the plantar surface (43.0%), the digital surface (22.5%), the dorsum (14.0%) or the apex (i.e. tip of toe, 11.8%). The duration of the index ulcer varied to a large degree, with a median of 1.84 months (range 3 days–12 years). A total of 72.0% of patients’ index ulcers were incident as opposed to recurrent. Only 3.5% of ulcers were solely ischaemic, 50.5% of ulcers were neuropathic only, and 45.5% of ulcers were both ischaemic and neuropathic.

Tables 9–11 summarise ulcer characterisation according to the PEDIS criteria, clinical signs and symptoms, and Wagner scale. Almost all patients (98%) had a grade 1 or 2 perfusion rating (no critical limb ischaemia); approximately equal proportions of patients had a grade 1–3 depth/tissue loss rating; the majority of patients (93.3%) had grade 2 sensation (loss of protective sensation); and the majority of patients had an infection of either grade 2 (inflammation of skin/subcutaneous tissue only, 37.3%) or grade 3 (extensive erythema deeper than skin/subcutaneous tissue, 59.3%). The majority of patients had an ulcer debridement undertaken at the baseline visit (87.8%), with the median area measuring 1.77 cm² (range 0.01–138.2 cm²). The clinical signs and symptoms classification of patients’ index ulcers revealed that 31.8% of patients had a foul wound odour; 42.5% had pocketing in the wound; 56.3% had discoloured granulation tissue; 51.0% had friable granulation tissue; 31.3% had a recent increase in pain, as opposed to the 2.3% who had a recent decrease in pain; 61.5% had a recent increase in wound size; and 31.5% had a breakdown of epithelium. Furthermore, of all index ulcers, 34.0% were classified as grade 1 (superficial diabetic ulcer); 33.5% were classified as grade 2 (ulcer extension to ligament, tendon, joint capsule or deep fascia without abscess or osteomyelitis); 30.5% were classified as grade 3 (deep ulcer with abscess, osteomyelitis or joint sepsis); 1.8% were classified as grade 4 (gangrene localised to a portion of forefoot or heel); and 0.3% were classified as grade 5 (extensive gangrenous involvement of the entire foot).

Tables 12 and 13 and Figure 5 summarise the antibiotic regimens patients were prescribed immediately pre and post sampling. Prior to sampling, 60.3% of patients had been treated with an antimicrobial dressing or agent on the infected ulcer. Furthermore, 46.8% of patients were on a systemic antibiotic regimen, with the most frequently prescribed antibiotics being flucloxacillin (31.1%), clindamycin (18.3%), co-amoxiclav (13.1%), ciprofloxacin (13.1%) and metronidazole (7.2%). The patient’s antibiotic regimen was changed following clinical assessment and specimen sampling, but before microbiology results were available, in 62.0% of patients. Among the 42.0% of patients who were not on an antibiotic regimen prior to sampling, treatment was initiated immediately post sampling. Finally, 6.5% of patients were not on an antibiotic regimen prior to sampling and did not have an antibiotic regimen initiated immediately post sampling.

FIGURE 5.

Prescribed antibiotics pre and post sampling (antibiotics are not mutually exclusive).

Coprimary end points

Coprimary end point: reported presence of likely pathogens

Most prevalent pathogens

Table 16 presents full cross-tabulations of the reported presence of the most prevalent pathogens (those with prevalence > 8%), Figure 6 depicts this information and Table 17 presents statistics relating to the agreement and differences in reporting of these pathogens.

TABLE 16 - Cross-tabulations of reported presence of at least one pathogen and most prevalent pathogens in order of taxonomic rank and prevalence

Pathogen (overall prevalence)	Swab results, n (%)		Tissue results, n (%)
At least one pathogen (88.1%)			Not reported	Reported	Total
	Swab	Not reported	47 (11.9)	71 (18.0)	118 (29.9)
		Reported	8 (2.0)	269 (68.1)	277 (70.1)
		Total	55 (13.9)	340 (86.1)	395 (100.0)
Gram-positive cocci (70.6%)			Not reported	Reported	Total
	Swab	Not reported	116 (29.4)	68 (17.2)	184 (46.6)
		Reported	14 (3.5)	197 (49.9)	211 (53.4)
		Total	130 (32.9)	265 (67.1)	395 (100.0)
Gram-negative bacilli (36.7%)			Not reported	Reported	Total
	Swab	Not reported	250 (63.3)	49 (12.4)	299 (75.7)
		Reported	12 (3.0)	84 (21.3)	96 (24.3)
		Total	262 (63.3)	133 (33.7)	395 (100.0)
Enterobacteriaceae (26.6%)			Not reported	Reported	Total
	Swab	Not reported	290 (73.4)	37 (9.4)	327 (82.8)
		Reported	14 (3.5)	54 (13.7)	68 (17.2)
		Total	304 (77.0)	91 (23.0)	395 (100.0)
Overall anaerobes (23.8%)			Not reported	Reported	Total
	Swab	Not reported	301 (76.2)	46 (11.6)	347 (87.8)
		Reported	19 (4.8)	29 (7.3)	48 (12.2)
		Total	320 (81.0)	75 (19.0)	395 (100.0)
Gram-positive bacilli (11.1%)			Not reported	Reported	Total
	Swab	Not Reported	351 (88.9)	40 (10.1)	391 (99.0)
		Reported	1 (0.3)	3 (0.8)	4 (1.0)
		Total	352 (89.1)	43 (10.9)	395 (100.0)
Streptococcus (16.7%)			Not reported	Reported	Total
	Swab	Not reported	329 (83.3)	18 (4.6)	347 (87.8)
		Reported	5 (1.3)	43 (10.9)	48 (12.2)
		Total	334 (84.6)	61 (15.4)	395 (100.0)
Enterococcus (excluding vancomycin resistant) (14.9%)			Not reported	Reported	Total
	Swab	Not reported	336 (85.1)	34 (8.6)	370 (93.7)
		Reported	6 (1.5)	19 (4.8)	25 (6.3)
		Total	342 (86.6)	53 (13.4)	395 (100.0)
CNS (12.2%)			Not reported	Reported	Total
	Swab	Not reported	347 (87.8)	39 (9.9)	386 (97.7)
		Reported	1 (0.3)	8 (2.0)	9 (2.3)
		Total	348 (88.1)	47 (11.9)	395 (100.0)
Corynebacterium (9.4%)			Not reported	Reported	Total
	Swab	Not reported	358 (90.6)	33 (8.4)	391 (99.0)
		Reported	1 (0.3)	3 (0.8)	4 (1.0)
		Total	359 (90.9)	36 (9.1)	395 (100.0)
Pseudomonas (8.6%)			Not reported	Reported	Total
	Swab	Not reported	361 (91.4)	8 (2.0)	369 (93.4)
		Reported	8 (2.0)	18 (4.6)	26 (6.6)
		Total	369 (93.4)	26 (6.6)	395 (100.0)
S. aureus (35.7%)			Not reported	Reported	Total
	Swab	Not reported	254 (64.3)	16 (4.1)	270 (68.4)
		Reported	16 (4.1)	109 (27.6)	125 (31.6)
		Total	270 (68.4)	125 (31.6)	395 (100.0)
MRSA (8.1%)			Not reported	Reported	Total
	Swab	Not reported	363 (91.9)	5 (1.3)	368 (93.2)
		Reported	1 (0.3)	26 (6.6)	27 (6.8)
		Total	364 (92.2)	31 (7.8)	395 (100.0)

FIGURE 6.

Reported presence of at least one pathogen and most prevalent pathogens. VRE, vanconycin-resistant Enterococcus.

TABLE 17 - Summary of agreement and disagreement statistics for most prevalent pathogens and the report of at least one pathogen

Tissue minus swab.

Exact p-value/CI.

Pathogens	Overall prevalence, %	Disagreement, %	Difference, % (95% CI)a	McNemar’s p-value	Agreement, %	κ (95% CI)	PABAK
≥ 1 pathogen	88.1	20.0	15.9 (11.8 to 20.1)	< 0.0001	80.0	0.44 (0.34 to 0.53)	0.60
Gram-positive cocci	70.6	20.8	13.7 (9.4 to 18.0)	< 0.0001	79.2	0.57 (0.50 to 0.65)	0.58
Gram-negative bacilli	36.7	15.4	9.4 (5.6 to 13.1)	< 0.0001	84.6	0.63 (0.55 to 0.71)	0.69
Enterobactereaceae (including coliforms)	26.6	12.9	5.8 (2.3 to 9.3)	0.0013	87.1	0.60 (0.50 to 0.70)	0.74
Overall anaerobes	23.8	16.5	6.8 (2.9 to 10.8)	0.0008	83.5	0.38 (0.26 to 0.50)	0.67
Gram-positive bacilli	11.1	10.4	9.9 (6.9 to 13.5)	< 0.0001b	89.6	0.11 (–0.01 to 0.23)	0.79
Streptococcus	16.7	5.8	3.3 (0.9 to 5.6)	0.0067	94.2	0.76 (0.66 to 0.85)	0.88
Enterococcus (excluding vancomycin resistant)	14.9	10.1	7.1 (4.0 to 10.1)	< 0.0001	89.9	0.44 (0.30 to 0.58)	0.80
CNS	12.2	10.1	9.6 (6.7 to 12.9)	< 0.0001b	89.9	0.26 (0.11 to 0.41)	0.80
Corynebacterium	9.4	8.6	8.1 (5.4 to 11.2)	< 0.0001b	91.4	0.13 (–0.01 to 0.28)	0.83
Pseudomonas	8.6	4.1	0.0 (–2.0 to 2.0)	1.0000	95.9	0.67 (0.52 to 0.82)	0.92
S. aureus	35.7	8.1	0.0 (–2.8 to 2.8)	1.0000	91.9	0.81 (0.75 to 0.87)	0.84
MRSA	8.1	1.5	1.0 (–0.2 to 2.8)	0.2188b	98.5	0.89 (0.80 to 0.98)	0.97

Overall, there was evidence of a significant difference [15.9% (95% CI 11.8% to 20.1%); p-value < 0.0001] between the swab and tissue samples in the percentage reporting at least one pathogen (86.1% of patients with tissue sample vs. 70.1% of patients with swab sample) (see Table 17).

Among the most prevalent pathogens, overall agreement between swab and tissue sample results was at least 79%. The κ-values for the chance corrected agreement suggested:

• almost perfect agreement for MRSA [κ = 0.89 (95% CI 0.80 to 0.98)] and S. aureus [κ = 0.81 (95% CI 0.75 to 0.87)]

• substantial agreement for Streptococcus [κ = 0.76 (95% CI 0.66 to 0.85)], Pseudomonas [κ = 0.67 (95% CI 0.52 to 0.82)] and Gram-negative bacilli [κ = 0.63 (95% CI 0.55 to 0.71)]

• moderate agreement for Enterobactereaceae (including coliforms) [κ = 0.60 (95% CI 0.50 to 0.70)], Gram-positive cocci [κ = 0.57 (95% CI 0.50 to 0.65)] and Enterococcus [κ = 0.44 (95% CI 0.30 to 0.58)]

• fair agreement for overall anaerobes [κ = 0.38 (95% CI 0.26 to 0.50)] and CNS [κ = 0.26 (95% CI 0.11 to 0.41)]

• slight agreement for Corynebacterium [κ = 0.13 (95% CI –0.01 to 0.28)] and Gram-positive bacilli [κ = 0.11 (95% CI –0.01 to 0.23)].

The PABAK for the majority of pathogens showed a considerably higher estimate of agreement after accounting for the low prevalence of the majority of pathogens.

For the majority of pathogens, there was evidence of a significant difference (McNemar’s p-value < 0.01), with reported prevalence higher in the tissue sample results than the swab results, with the exception for S. aureus, MRSA and Pseudomonas. Symmetrical disagreement was observed for S. aureus and Pseudomonas, with the pathogen reported in one sample but not the other an equal number of times for the two samples. The reported prevalence of MRSA was non-statistically higher in tissue samples than swab samples.

Semi-quantitative extent of bacterial growth

Table 18 presents cross-tabulations of the level of growth of each of the most prevalent pathogens, by swab and tissue samples, and the associated κ-values. The overall κ-value does not account for the ordinal levels of growth, whereas the weighted κ-value quantifies the relative difference between levels of growth. κ-values were calculated after excluding patients with a missing level of growth for either the swab or tissue sample.

TABLE 18 - Cross-tabulations on the semiquantitative extent of bacterial growth and κ-statistics

Weighted 1, exponential values; weighted 2, linear values (sensitivity analysis).

Tissue results: level of growth						Total, n (%)	κ-valuea	κ	95% CI
	Not reported	Reported: no growth	+	++	+++
Gram-positive cocci
Swab results, n (%)							(n = 340)
Not reported	116 (29.4)	13 (3.3)	25 (6.3)	14 (3.5)	16 (4.1)	184 (46.6)	Overall	0.49	0.42 to 0.55
Reported: no growth	3 (0.8)	33 (8.4)	1 (0.3)	0 (0.0)	1 (0.3)	38 (9.6)	Weighted 1	0.58	0.51 to 0.66
+	4 (1.0)	1 (0.3)	22 (5.6)	3 (0.8)	0 (0.0)	30 (7.6)	Weighted 2	0.60	0.54 to 0.67
++	3 (0.8)	0 (0.0)	13 (3.3)	15 (3.8)	10 (2.5)	41 (10.4)
+++	4 (1.0)	3 (0.8)	9 (2.3)	21 (5.3)	65 (16.5)	102 (25.8)
Total	130 (32.9)	50 (12.7)	70 (17.7)	53 (13.4)	92 (23.3)	n = 395
Gram-negative bacilli
Swab results, n (%)							(n = 354)
Not reported	250 (63.3)	5 (1.3)	14 (3.5)	12 (3.0)	18 (4.6)	299 (75.7)	Overall	0.48	0.39 to 0.57
Reported: no growth	4 (1.0)	18 (4.6)	2 (0.5)	3 (0.8)	5 (1.3)	32 (8.1)	Weighted 1	0.49	0.38 to 0.61
+	2 (0.5)	1 (0.3)	6 (1.5)	0 (0.0)	2 (0.5)	11 (2.8)	Weighted 2	0.53	0.43 to 0.63
++	3 (0.8)	2 (0.5)	4 (1.0)	10 (2.5)	2 (0.5)	21 (5.3)
+++	3 (0.8)	1 (0.3)	5 (1.3)	5 (1.3)	18 (4.6)	32 (8.1)
Total	262 (66.3)	27 (6.8)	31 (7.8)	30 (7.6)	45 (11.4)	n = 395
Enterobacteriaceae (including coliforms)
Swab results, n (%)							(n = 359)
Not reported	290 (73.4)	6 (1.5)	12 (3.0)	5 (1.3)	14 (3.5)	327 (82.8)	Overall	0.45	0.33 to 0.56
Reported: no growth	4 (1.0)	17 (4.3)	2 (0.5)	1 (0.3)	3 (0.8)	27 (6.8)	Weighted 1	0.50	0.35 to 0.64
+	2 (0.5)	1 (0.3)	2 (0.5)	0 (0.0)	0 (0.0)	5 (1.3)	Weighted 2	0.51	0.38 to 0.64
++	3 (0.8)	1 (0.3)	3 (0.8)	5 (1.3)	0 (0.0)	12 (3.0)
+++	5 (1.3)	1 (0.3)	2 (0.5)	3 (0.8)	13 (3.3)	24 (6.1)
Total	304 (77.0)	26 (6.6)	21 (5.3)	14 (3.5)	30 (7.6)	n = 395
Overall anaerobes
Swab results, n (%)							(n = 377)
Not reported	301 (76.2)	9 (2.3)	12 (3.0)	14 (3.5)	11 (2.8)	347 (87.8)	Overall	0.32	0.21 to 0.43
Reported: no growth	4 (1.0)	4 (1.0)	0 (0.0)	0 (0.0)	0 (0.0)	8 (2.0)	Weighted 1	0.43	0.29 to 0.57
+	7 (1.8)	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.3)	8 (2.0)	Weighted 2	0.44	0.31 to 0.56
++	1 (0.3)	1 (0.3)	1 (0.3)	4 (1.0)	6 (1.5)	13 (3.3)
+++	7 (1.8)	0 (0.0)	0 (0.0)	2 (0.5)	10 (2.5)	19 (4.8)
Total	320 (81.0)	14 (3.5)	13 (3.3)	20 (5.1)	28 (7.1)	n = 395
Gram-positive bacilli
Swab results, n (%)							(n = 389)
Not reported	351 (88.9)	6 (1.5)	15 (3.8)	10 (2.5)	9 (2.3)	391 (99.0)	Overall	0.06	–0.00 to 0.13
+	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	Weighted 1	0.07	–0.01 to 0.16
++	1 (0.3)	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.3)	2 (0.5)	Weighted 2	0.11	–0.01 to 0.24
+++	0 (0.0)	0 (0.0)	1 (0.3)	1 (0.3)	0 (0.0)	2 (0.5)
Total	352 (89.1)	6 (1.5)	16 (4.1)	11 (2.8)	10 (2.5)	n = 395
Streptococcus
Swab results, n (%)							(n = 384)
Not reported	329 (83.3)	6 (1.5)	5 (1.3)	1 (0.3)	6 (1.5)	347 (87.8)	Overall	0.65	0.55 to 0.75
Reported: no growth	0 (0.0)	4 (1.0)	0 (0.0)	0 (0.0)	1 (0.3)	5 (1.3)	Weighted 1	0.68	0.56 to 0.80
+	2 (0.5)	0 (0.0)	5 (1.3)	1 (0.3)	0 (0.0)	8 (2.0)	Weighted 2	0.74	0.65 to 0.84
++	1 (0.3)	0 (0.0)	1 (0.3)	4 (1.0)	2 (0.5)	8 (2.0)
+++	2 (0.5)	0 (0.0)	2 (0.5)	7 (1.8)	16 (4.1)	27 (6.8)
Total	334 (84.6)	10 (2.5)	13 (3.3)	13 (3.3)	25 (6.3)	n = 395
Enterococcus (excluding vancomycin resistant)
Swab results, n (%)							(n = 384)
Not reported	336 (85.1)	6 (1.5)	9 (2.3)	12 (3.0)	7 (1.8)	370 (93.7)	Overall	0.39	0.24 to 0.54
Reported – no growth	0 (0.0)	4 (1.0)	0 (0.0)	0 (0.0)	0 (0.0)	4 (1.0)	Weighted 1	0.52	0.34 to 0.70
+	2 (0.5)	0 (0.0)	1 (0.3)	0 (0.0)	0 (0.0)	3 (0.8)	Weighted 2	0.47	0.31 to 0.64
++	1 (0.3)	1 (0.3)	0 (0.0)	3 (0.8)	1 (0.3)	6 (1.5)
+++	3 (0.8)	0 (0.0)	0 (0.0)	1 (0.3)	8 (2.0)	12 (3.0)
Total	342 (86.6)	11 (2.8)	10 (2.5)	16 (4.1)	16 (4.1)	n = 395
CNS
Swab results, n (%)							(n = 389)
Not reported	347 (87.8)	5 (1.3)	22 (5.6)	7 (1.8)	5 (1.3)	386 (97.7)	Overall	0.23	0.08 to 0.37
+	0 (0.0)	0 (0.0)	2 (0.5)	0 (0.0)	0 (0.0)	2 (0.5)	Weighted 1	0.34	0.11 to 0.57
++	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.3)	2 (0.5)	3 (0.8)	Weighted 2	0.31	0.13 to 0.50
+++	1 (0.3)	1 (0.3)	0 (0.0)	0 (0.0)	2 (0.5)	4 (1.0)
Total	348 (88.1)	6 (1.5)	24 (6.1)	8 (2.0)	9 (2.3)	n = 395
Corynebacterium
Swab results, n (%)							(n = 390)
Not reported	358 (90.6)	5 (1.3)	10 (2.5)	10 (2.5)	8 (2.0)	391 (99.0)	Overall	0.07	–0.00 to 0.15
++	1 (0.3)	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.3)	2 (0.5)	Weighted 1	0.08	–0.01 to 0.17
+++	0 (0.0)	0 (0.0)	1 (0.3)	1 (0.3)	0 (0.0)	2 (0.5)	Weighted 2	0.13	–0.01 to 0.27
Total	359 (90.9)	5 (1.3)	11 (2.8)	11 (2.8)	9 (2.3)	n = 395
Pseudomonas
Swab results, n (%)							(n = 391)
Not reported	361 (91.4)	1 (0.3)	1 (0.3)	3 (0.8)	3 (0.8)	369 (93.4)	Overall	0.58	0.42 to 0.75
Reported: no growth	0 (0.0)	2 (0.5)	0 (0.0)	0 (0.0)	1 (0.3)	3 (0.8)	Weighted 1	0.61	0.40 to 0.82
+	4 (1.0)	0 (0.0)	3 (0.8)	0 (0.0)	1 (0.3)	8 (2.0)	Weighted 2	0.61	0.44 to 0.79
++	2 (0.5)	0 (0.0)	1 (0.3)	3 (0.8)	0 (0.0)	6 (1.5)
+++	2 (0.5)	0 (0.0)	1 (0.3)	0 (0.0)	6 (1.5)	9 (2.3)
Total	369 (93.4)	3 (0.8)	6 (1.5)	6 (1.5)	11 (2.8)	n = 395
S. aureus
Swab results, n (%)							(n = 363)
Not reported	254 (64.3)	1 (0.3)	11 (2.8)	3 (0.8)	1 (0.3)	270 (68.4)	Overall	0.64	0.57 to 0.71
Reported: no growth	4 (1.0)	25 (6.3)	0 (0.0)	0 (0.0)	0 (0.0)	29 (7.3)	Weighted 1	0.7	0.62 to 0.79
+	3 (0.8)	1 (0.3)	14 (3.5)	1 (0.3)	0 (0.0)	19 (4.8)	Weighted 2	0.74	0.68 to 0.81
++	6 (1.5)	0 (0.0)	8 (2.0)	8 (2.0)	2 (0.5)	24 (6.1)
+++	3 (0.8)	1 (0.3)	9 (2.3)	10 (2.5)	30 (7.6)	53 (13.4)
Total	270 (68.4)	28 (7.1)	42 (10.6)	22 (5.6)	33 (8.4)	n = 395
MRSA
Swab results, n (%)							(n = 391)
Not reported	363 (91.9)	0 (0.0)	1 (0.3)	2 (0.5)	2 (0.5)	368 (93.2)	Overall	0.73	0.61 to 0.85
Reported: no growth	0 (0.0)	4 (1.0)	0 (0.0)	0 (0.0)	0 (0.0)	4 (1.0)	Weighted 1	0.79	0.67 to 0.91
+	1 (0.3)	0 (0.0)	3 (0.8)	0 (0.0)	0 (0.0)	4 (1.0)	Weighted 2	0.83	0.73 to 0.93
++	0 (0.0)	0 (0.0)	2 (0.5)	1 (0.3)	2 (0.5)	5 (1.3)
+++	0 (0.0)	0 (0.0)	0 (0.0)	3 (0.8)	11 (2.8)	14 (3.5)
Total	364 (92.2)	4 (1.0)	6 (1.5)	6 (1.5)	15 (3.8)	n = 395

Agreement on the level of growth (according to the primary weighting) was somewhat skewed owing to the prevalence of each pathogen and the proportion of patients with discordant results where a pathogen was reported in one sample and not the other. Therefore, the level of growth in one sample was often in comparison with the lack of presence of the pathogen rather than a corresponding level of growth. The range for agreement was substantial for Streptococcus, Pseudomonas, S. aureus and MRSA; moderate for Gram-positive cocci, Gram-negative bacilli, Enterobacteriaceae and Enterococcus; fair for CNS and Corynebacterium; and slight for Gram-positive bacilli.

Summary of pathogens

Table 19 presents the overall summary of all pathogens reported by specimen type for the evaluable population and by baseline characteristics. Figure 7 presents the overall summary by centre. Overall, there is a difference in the pathogens reported by the two techniques for 58.0% of patients. Findings among the 395 patient pairs of results were swab and tissue results reported the same pathogens in 42.0% of patients; swab results reported additional pathogens to those in the tissue in 8.1% of patients; tissue reported additional pathogens to those in the swab in 36.7% of patients; and the tissue sample and swab specimens reported different pathogens, with or without overlap, in 13.2% of patients.

TABLE 19 - Overall summary of pathogens by baseline characteristics

The swab and tissue samples reported the same pathogens for 47 (11.9%) patients as no pathogens were reported for either sample.

The swab sample reported additional pathogens for 8 (2.0%) patients as no pathogens were reported from the tissue sample only.

The tissue sample reported additional pathogens for 71 (18.0%) patients as no pathogens were reported from the swab only.

With or without overlap.

Baseline characteristics	Swab and tissue report the same pathogens, n (%)	Swab reports additional pathogens to the tissue, n (%)	Tissue reports additional pathogens to the swab, n (%)	Swab and tissue report different pathogens,d n (%)
Total (n = 395)	166 (42.0)a	32 (8.1)b	145 (36.7)c	52 (13.2)
Type of ulcer
Any ischaemia (± neuropathy) (n = 194)	87 (44.8)	17 (8.8)	68 (35.1)	22 (11.3)
Neuropathic only (n = 199)	79 (39.7)	15 (7.5)	76 (38.2)	29 (14.6)
Missing (n = 2)	0 (0.0)	0 (0.0)	1 (50.0)	1 (50.0)
Grade of ulcer
Grade 1 (n = 135)	60 (44.4)	12 (8.9)	47 (34.8)	16 (11.9)
Grade 2 (n = 132)	59 (44.7)	8 (6.1)	48 (36.4)	17 (12.9)
Grade 3, 4 or 5 (n = 128)	47 (36.7)	12 (9.4)	50 (39.1)	19 (14.8)
Pre-sampling antibiotic therapy
Yes (n = 186)	78 (41.9)	12 (6.5)	71 (38.2)	25 (13.4)
No (n = 190)	80 (42.1)	15 (7.9)	70 (36.8)	25 (13.2)
Missing (n = 19)	8 (42.1)	5 (26.3)	4 (21.1)	2 (10.5)
Presence of antimicrobial dressing or agent
Yes (n = 238)	101 (42.4)	20 (8.4)	82 (34.5)	35 (14.7)
No (n = 152)	61 (40.1)	11 (7.2)	63 (41.4)	17 (11.2)
Missing (n = 5)	4 (80.0)	1 (20.0)	0 (0.0)	0 (0.0)
Wound duration (by median)
< 56 days (n = 189)	72 (38.1)	13 (6.9)	81 (42.9)	23 (12.2)
≥ 56 days (n = 202)	94 (46.5)	18 (8.9)	62 (30.7)	28 (13.9)
Missing (n = 4)	0 (0.0)	1 (25.0)	2 (50.0)	1 (25.0)

FIGURE 7.

Overall summary of all pathogens reported by centre. SJUH, St James’s University Hospital.

Multinomial regression analyses

Multinomial regression modelling with a random effect for centre (and MIs to allow for missing data) was used to assess whether or not agreement, based on the overall summary of pathogens, was influenced by the pre-specified baseline covariates (Table 20).

TABLE 20 - Multinomial regression analyses for individually fitted baseline factors on the overall summary of pathogens with random-centre effect

AIC, Akaike information criterion; df, degrees of freedom.

An odds ratio of > 1.0 indicates an increase in the odds of the specified outcome compared with the outcome reference (both swab and tissue report the same pathogens) for the specified factor compared with its reference group (listed after vs. in the table).

Smaller is better.

Based on the reduction in –2 log-likelihood from null model.

Multinomial regressionanalyses	Summary of pathogens (reference: both swab and tissue report the same pathogens)	Odds ratioa (95% CI)	AICb	df	p-valuec
Null model			941.29
Ulcer type: any ischaemia (± neuropathy) vs. neuropathic only			945.72	3	0.6663
	Swab reports additional pathogens to the tissue	1.03 (0.48 to 2.20)
	Tissue reports additional pathogens to the swab	0.86 (0.53 to 1.40)
	Both swab and tissue report different pathogens	0.68 (0.35 to 1.31)
Ulcer grade			949.16	6	0.6598
Grade 2 vs. grade 1	Swab reports additional pathogens to the tissue	0.68 (0.26 to 1.78)
	Tissue reports additional pathogens to the swab	1.08 (0.60 to 1.93)
	Both swab and tissue report different pathogens	1.14 (0.51 to 2.54)
Grade 3/4/5 vs. grade 1	Swab reports additional pathogens to the tissue	1.28 (0.52 to 3.11)
	Tissue reports additional pathogens to the swab	1.60 (0.87 to 2.95)
	Both swab and tissue report different pathogens	1.55 (0.69 to 3.45)
Pre-sampling antibiotic therapy: yes vs. no			946.28	3	0.8001
	Swab reports additional pathogens to the tissue	0.80 (0.36 to 1.80)
	Tissue reports additional pathogens to the swab	1.14 (0.69 to 1.89)
	Both swab and tissue report different pathogens	1.10 (0.56 to 2.16)
Antimicrobial dressing: yes vs. no			943.44	3	0.2782
	Swab reports additional pathogens to the tissue	1.13 (0.51 to 2.51)
	Tissue reports additional pathogens to the swab	0.69 (0.40 to 1.19)
	Both swab and tissue report different pathogens	1.38 (0.66 to 2.89)
Wound duration (median split): ≥ 56 days vs. < 56 days			941.48	3	0.1216
	Swab reports additional pathogens to the tissue	1.06 (0.49 to 2.32)
	Tissue reports additional pathogens to the swab	0.57 (0.35 to 0.93)
	Both swab and tissue report different pathogens	0.88 (0.46 to 1.70)
Log-wound duration (continuous)			944.97	3	0.5091
	Swab reports additional pathogens to the tissue	0.95 (0.72 to 1.25)
	Tissue reports additional pathogens to the swab	0.88 (0.74 to 1.04)
	Both swab and tissue report different pathogens	0.93 (0.74 to 1.18)

None of the baseline factors [ulcer type (any ischaemia/neuropathic only), ulcer grade (Wagner grade 1/grade 2/grade 3, 4 or 5), pre-sampling antibiotic therapy (yes/no), antimicrobial dressing or agent (yes/no), wound duration (considered dichotomously as < 56 days/≥ 56 days and continuously on the log-scale)] was found to have a significant overall impact on agreement. However, comparison of the individual outcomes did suggest that patients with a wound duration ≥ 56 days had significantly reduced odds of their tissue sample reporting additional pathogens to the swab sample, as opposed to their swab and tissue reporting the same pathogens, with an odds ratio of 0.57 (95% CI 0.35 to 0.93). This finding was not, however, supported on the continuous scale for wound duration.

Coprimary end point 2: reported presence of antimicrobial resistance among likely pathogens

Likely pathogens

Of the three pathogens of interest, no meticillin-resistant CNS was reported and vancomycin-resistant Enterococcus was reported for just one patient in both their swab and tissue sample results (Table 21).

TABLE 21 - Reported presence of antimicrobial resistance among likely pathogens

Antimicrobial resistance	Swab (N = 395), n (%)	Tissue (N = 395), n (%)	Overall prevalence (N = 395), n (%)
MRSA	27 (6.8)	31 (7.8)	32 (8.1)
Vancomycin-resistantEnterococcus	1 (0.3)	1 (0.3)	1 (0.3)
Meticillin-resistant CNS	0 (0)	0 (0)	0 (0)

Meticillin-resistant S. aureus was reported in 32 (8.1%) patients overall, with overall agreement of 98.5% between swab and tissue samples. In 5 (1.3%) patients, the pathogen was reported in the tissue results but not the swab, and in 1 (0.3%) patient, the pathogen was reported in the swab but not the tissue results (see Table 16). As such, a difference of 1.0% (exact 95% CI –0.2% to 2.8%) was reported, with McNemar’s test suggesting that this was not a significant difference (exact p-value = 0.2188) (see Table 17).

To evaluate whether or not agreement was influenced by pre-specified covariates, multinomial regression modelling had been proposed based on the outcomes of reported MRSA: by swab not tissue, by tissue not swab, swab and tissue results agree. However, given the small number of patients whose swab and tissue sample results did not agree [6 (1.6%)], this analysis was not appropriate and was not performed.

Additional sensitivities and resistances

In addition to the three pathogens of interest, resistance and sensitivity to antibiotics were collected where reported for all pathogens within a patient’s swab or tissue sample.

Patients’ swab or tissue sample results were reported to contain pathogens with a resistance to a maximum of eight different antibiotics and sensitivity to a maximum of 10 different antibiotic agents, of any of the antibiotics for which samples were tested (Table 22). There were 123 (31.1%) patients whose swab sample reported pathogen(s) with resistance to at least one antibiotic agent, whereas 165 (41.8%) patients’ tissue samples reported pathogen(s) with resistance to at least one resistant antibiotic agent. Overall, from either the swab or tissue sample results there were 185 (46.8%) patients for whom resistance to at least one antibiotic agent was reported. A greater proportion of patients’ sample results reported at least one antibiotic to which pathogens were sensitive. There were 221 (55.9%) patients whose swab samples reported pathogen(s) with sensitivity to at least one antibiotic agent, 268 (67.8%) patients whose tissue sample reported pathogen(s) with sensitivity to at least one antibiotic agent and 284 (71.9%) patients for whom the swab or tissue sample reported pathogen(s) with sensitivity to at least one antibiotic agent.

TABLE 22 - Summary of the number of different antibiotic agents for which pathogens within swab and tissue sample results were found to be resistant or sensitive

IQR, interquartile range; SD, standard deviation.

Number of antibiotics	Antibiotic resistance			Antibiotic sensitivity
	Swab (N = 395)	Tissue (N = 395)	Overall (N = 395)	Swab (N = 395)	Tissue (N = 395)	Overall (N = 395)
Number of antibiotics
Mean (SD)	0.6 (1.16)	0.9 (1.35)	1.0 (1.45)	1.6 (1.80)	2.1 (1.98)	2.5 (2.14)
Median (range)	0.0 (0–6)	0.0 (0–8)	0.0 (0–8)	1.0 (0–9)	2.0 (0–9)	2.0 (0–10)
IQR	(0.0–1.0)	(0.0–1.0)	(0.0–2.0)	(0.0–3.0)	(0.0–3.0)	(0.0–4.0)
Missing	0	0	0	0	0	0
Number of antibiotics, n (%)
0	272 (68.9)	230 (58.2)	210 (53.2)	174 (44.1)	127 (32.2)	111 (28.1)
1	62 (15.7)	77 (19.5)	81 (20.5)	38 (9.6)	39 (9.9)	27 (6.8)
2	31 (7.8)	41 (10.4)	51 (12.9)	63 (15.9)	69 (17.5)	64 (16.2)
3	10 (2.5)	21 (5.3)	19 (4.8)	56 (14.2)	68 (17.2)	68 (17.2)
4	14 (3.5)	17 (4.3)	21 (5.3)	35 (8.9)	50 (12.7)	58 (14.7)
5	3 (0.8)	4 (1.0)	6 (1.5)	16 (4.1)	14 (3.5)	28 (7.1)
6	3 (0.8)	4 (1.0)	5 (1.3)	10 (2.5)	18 (4.6)	23 (5.8)
7	0 (0.0)	0 (0.0)	1 (0.3)	2 (0.5)	5 (1.3)	10 (2.5)
8	0 (0.0)	1 (0.3)	1 (0.3)	0 (0.0)	4 (1.0)	4 (1.0)
9	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.3)	1 (0.3)	1 (0.3)
10	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	1 (0.3)

The most frequently reported antibiotic agent to which at least one pathogen isolated from a patient’s swab or tissue sample was resistant was penicillin, with a resistance observed in either the swab or tissue sample results for 72 (18.2%) patients; that is, 47 (11.9%) patients’ swab samples and 63 (15.9%) patients’ tissue samples. Figure 8 presents all reported antibiotics to which pathogens were found to be resistant within patients’ samples. A similar pattern is observed across all reported antibiotic agents with each reported more often in the tissue sample than in the swab.

FIGURE 8.

Reported antibiotic resistances for pathogens within patients swab and tissue sample results.

The most frequently reported antibiotic agent to which at least one pathogen isolated from a patient’s swab or tissue sample was sensitive was flucloxacillin, with a sensitivity observed in either the swab or tissue sample results for 144 (36.5%) patients; that is, 126 (31.9%) patients’ swab samples and 126 (31.9%) patients’ tissue samples. Figure 9 presents all reported antibiotics to which pathogens were found to be resistant within patients’ samples, with a similar pattern observed across the majority of reported antibiotic agents, with all agents but erythromycin reported in the same or a greater percentage of patients in the tissue sample than the swab sample.

FIGURE 9.

Reported antibiotic sensitivities for pathogens within patients swab and tissue sample results.

Coprimary end point 3: number of pathogens reported per specimen

The third coprimary end point evaluated agreement between the two specimen collection methods for microbiological characterisation determined by the number of pathogens reported per specimen.

Tables 23 and 24 present the cross-tabulation and summary statistics of the number of pathogens reported from each sample. A median of 1 pathogen was reported in both samples, and the mean number of pathogens reported in the swab and tissue samples was 1 and 1.5, respectively, with a slightly higher level of variation observed in the tissue samples. The number of pathogens ranged from 0 to 4 in the swab sample and 0 to 6 in the tissue sample. A greater proportion of swab results reported no pathogens compared with tissue results (29.9% vs. 13.9%), whereas similar proportions of samples reported just one pathogen (45.1% and 40.8%). Where more than one pathogen was reported, there was consistently a greater frequency of patients with more pathogens in the tissue sample than the swab sample results.

TABLE 23 - Cross-tabulation of the number of pathogens reported per specimen

Green text indicates agreement.

Swab results	Tissue results, n (%)					Total
	0	1	2	3	4 or more
0	47 (11.9)	44 (11.1)	16 (4.1)	10 (2.5)	1 (0.3)	118 (29.9)
1	7 (1.8)	96 (24.3)	50 (12.7)	17 (4.3)	8 (2.0)	178 (45.1)
2	1 (0.3)	20 (5.1)	43 (10.9)	13 (3.3)	4 (1.0)	81 (20.5)
3	0 (0.0)	1 (0.3)	6 (1.5)	8 (2.0)	1 (0.3)	16 (4.1)
4 or more	0 (0.0)	0 (0.0)	0 (0.0)	0 (0.0)	2 (0.5)	2 (0.5)
Total	55 (13.9)	161 (40.8)	115 (29.1)	48 (12.2)	16 (4.1)	395 (100.0)

TABLE 24 - Summary statistics of the number of pathogens reported per specimen

SD, standard deviation.

Number of pathogens reported per specimen	Swab (N = 395)	Tissue (N = 395)
Mean (SD)	1.0 (0.84)	1.5 (1.04)
Median (range)	1.0 (0–4)	1.0 (0–6)
Range	0–4	0–6
Number of pathogens: frequency, n (%)
0	118 (29.9)	55 (13.9)
1	178 (45.1)	161 (40.8)
2	81 (20.5)	115 (29.1)
3	16 (4.1)	48 (12.2)
4	2 (0.5)	13 (3.3)
5	0 (0.0)	2 (0.5)
6	0 (0.0)	1 (0.3)

A summary of the number of pathogens by baseline characteristics, presented in Table 25, shows that for approximately half (49.6%) of all patients the same number of pathogens were reported for the tissue and swab sample; for 41.5% of patients the tissue sample reported at least one more pathogen than the swab; and for 8.9% of patients the swab sample reported at least one more pathogen than the tissue. Figure 10 presents the summary of the number of pathogens by centre.

TABLE 25 - Summary of the number of pathogens by baseline characteristics

Baseline characteristics	1: swab sampling had ≥ 2 extra pathogens reported	2: swab sampling had 1 extra pathogen reported	3: tissue and swab sampling had the same number of pathogens reported	4: tissue sampling had 1 extra pathogen reported	5: tissue sampling had ≥ 2 extra pathogens reported
Total (N = 395), n (%)	2 (0.5)	33 (8.4)	196 (49.6)	108 (27.3)	56 (14.2)
Type of ulcer, n (%)
Any ischaemia (± neuropathy) (n = 194)	0 (0.0)	18 (9.3)	101 (52.1)	50 (25.8)	25 (12.9)
Neuropathic only (n = 199)	2 (1.0)	15 (7.5)	94 (47.2)	57 (28.6)	31 (15.6)
Missing (n = 2)	0 (0.0)	0 (0.0)	1 (50.0)	1 (50.0)	0 (0.0)
Grade of ulcer, n (%)
Grade 1 (n = 135)	1 (0.7)	13 (9.6)	69 (51.1)	33 (24.4)	19 (14.1)
Grade 2 (n = 132)	0 (0.0)	8 (6.1)	68 (51.5)	35 (26.5)	21 (15.9)
Grade 3, 4 or 5 (n = 128)	1 (0.8)	12 (9.4)	59 (46.1)	40 (31.3)	16 (12.5)
Pre-sampling antibiotic therapy, n (%)
Yes (n = 186)	0 (0.0)	15 (8.1)	92 (49.5)	51 (27.4)	28 (15.1)
No (n = 190)	2 (1.1)	13 (6.8)	94 (49.5)	54 (28.4)	27 (14.2)
Missing (n = 19)	0 (0.0)	5 (26.3)	10 (52.6)	3 (15.8)	1 (5.3)
Presence of antimicrobial dressing or agent, n (%)
Yes (n = 238)	0 (0.0)	23 (9.7)	123 (51.7)	58 (24.4)	34 (14.3)
No (n = 152)	2 (1.3)	9 (5.9)	69 (45.4)	50 (32.9)	22 (14.5)
Missing (n = 5)	0 (0.0)	1 (20.0)	4 (80.0)	0 (0.0)	0 (0.0)
Wound duration, n (%)
< 56 days (n = 189)	1 (0.5)	13 (6.9)	85 (45.0)	57 (30.2)	33 (17.5)
≥ 56 days (n = 202)	1 (0.5)	19 (9.4)	110 (54.5)	49 (24.3)	23 (11.4)
Missing (n = 4)	0 (0.0)	1 (25.0)	1 (25.0)	2 (50.0)	0 (0.0)

FIGURE 10.

Summary of the number of pathogens reported by centre. SJUH, St James’s University Hospital.

Ordinal regression analyses

Ordinal regression modelling with a random effect for centre (and MIs to allow for missing data) was used to assess whether or not agreement, based on the summary of the number of pathogens, was influenced by the pre-specified baseline covariates. The results are presented in Table 26.

TABLE 26 - Ordinal regression analyses for individually fitted baseline factors on the summary of the number of pathogens with random-centre effect

AIC, Akaike information criterion; df, degrees of freedom.

An odds ratio of > 1.0 indicating a positive relationship with outcome compared with the variable reference group (those listed after ‘vs.’ in the table). For example, an odds ratio > 1.0 indicates a tendency towards higher levels of the ordinal outcome (i.e. in the direction that the tissue sample finds 2 or more pathogens) compared with the variable reference group.

Smaller is better.

Based on the reduction in –2 log-likelihood from null model.

Significant at the 5% level.

Baseline characteristics	Odds ratioa (95% CI)	AICb	df	p-valuec
Null model		917.72
Ulcer type: any ischaemia (± neuropathy) vs. neuropathic only	0.90 (0.61 to 1.33)	919.45	1	0.6030
Ulcer grade		920.16	2	0.4587
Grade 2 vs. grade 1	1.33 (0.82 to 2.15)
Grade 3, 4 or 5 vs. grade 1	1.27 (0.78 to 2.07)
Pre-sampling antibiotic therapy: yes vs. no	1.25 (0.81 to 1.91)	918.56	1	0.2828
Antimicrobial dressing: yes vs. no	0.76 (0.49 to 1.18)	918.16	1	0.2127
Wound duration (median split): ≥ 56 days vs. < 56 days	0.64 (0.43 to 0.95)	914.62	1	0.0240d
Log-wound duration (continuous)	0.92 (0.80 to 1.05)	918.15	1	0.2101

Of the baseline factors [ulcer type (any ischaemia/neuropathic only), ulcer grade (Wagner grade 1/grade 2/grade 3, 4 or 5), pre-sampling antibiotic therapy (yes/no), antimicrobial dressing or agent (yes/no), wound duration (considered dichotomously as < 56 days/≥ 56 days and continuously on the log-scale)], only wound duration (< 56 days/≥ 56 days) was found to have a statistically significant association (p-value = 0.0240). The associated odds ratio of 0.64 (95% CI 0.43 to 0.95) suggests that patients whose ulcer has been present for 56 days or more had significantly reduced odds of having a higher outcome (i.e. in the direction that the tissue sampling had two or more extra pathogens) than those whose ulcer has been present for fewer than 56 days.

Owing to the significance of wound duration, a forward selection model building approach was used to determine if further covariates had an influential effect on outcome in the model containing wound duration and centre random effect. Table 27 presents the results of the model building. There was no further significant improvement in the fit of the model on the addition of any additional baseline factors, and so the final model contained wound duration and random-centre effect and is presented in Table 28 and Figure 11.

TABLE 27 - Sequential chi-squared tests for the reduction in –2 log-likelihood: wound duration fixed effect (1 degree of freedom) and centre random effect (1 degree of freedom)

AIC, Akaike information criterion; df, degrees of freedom.

Smaller is better.

Based on the reduction in –2 log-likelihood from the model containing wound duration and centre random effect.

Additional baseline characteristic	Reduction in df	AICa	Reduction in –2log-likelihood	p-valueb
Wound duration (median split)		914.62
+ Ulcer type	1	916.53	0.086	0.7688
+ Ulcer grade	2	916.69	1.925	0.3820
+ Pre-sampling antibiotic therapy	1	915.94	0.674	0.4115
+ Antimicrobial dressing	1	915.60	1.019	0.3128

TABLE 28 - Final ordinal logistic regression model containing random-centre effect and fixed effect for wound duration

H0, null hypothesis.

Fixed effect	Odds ratio (95% CI)	Estimates of fixed effect			Random-centre effect
		Parameter estimate	Standard error	p-value	Parameter estimate	Standard error	Test of H0: random intercept variance = 0
Wound duration: ≥ 56 days vs. < 56 days	0.64 (0.43 to 0.95)	–0.4501	0.2008	0.0250	0.4689	0.2065	< 0.0001

FIGURE 11.

Plot of predicted random-centre effect in the model with fixed effect for wound duration (median split). SJUH, St James’s University Hospital.

Graphical plots were used to assess the proportional odds assumption for each baseline factor and can be found in Appendix 1. The proportional odds assumption was supported for all factors with the exception of centre, which was, however, fitted as a random effect negating the need for proportional odds.

The figure presents the ranked predicted random-centre effect on the parameter estimate of wound duration of –0.4501 (see Table 28). The parameter estimate relates to the odds of a higher outcome (i.e. tissue sample finds two or more pathogens), with a negative value reducing these odds, and a positive increasing the odds. The presented predicted centre effects vary considerably, across both positive and negative values, and therefore impact on the likely odds of a higher outcome and variability around the estimate across centres.

Discussion

This study is a cross-sectional multicentre study to examine agreement and disagreement between swab and tissue sampling techniques in patients with a suspected DFU infection. The conclusions drawn from this study will help to determine if the extra effort and cost of sampling tissue is potentially worthwhile. Furthermore, if there is disagreement, we aimed to determine whether one method provided additional information or different information.

The key results were that a significant proportion of wounds suspected to be infected had microbiology reports that indicated no growth, and a further proportion indicated no pathogens. There was a higher proportion of swab samples than tissue samples that had no reported pathogens. There are a number of possible explanations why the culture results may report no pathogens, such as the clinical diagnosis being incorrect (e.g. inflammation was mistaken for infection). Given the lack of validated tools for diagnosing the presence of wound infection and the acknowledged risk of missing infection in diabetic foot ulceration, it is understandable that clinical diagnosis might prioritise sensitivity over specificity (accepting practice that misclassifies people as having an infected ulcer when it is not, but not missing anyone with an infected ulcer). Furthermore, it may be that sampling technique as currently practised in these centres may not have adequately captured wound flora. For example, if there was inadequate wound ulcer debridement, then swabbing or taking a tissue sample from a sloughy ulcer area will be likely to collect surface contaminants, with or without wound tissue bacteria. Last, the lack of any organisms identified from a sample may be attributable to them not having survived the transport to the microbiology laboratory. Our information from sites indicated a range of transport media, and in some centres dry tissue samples were transported, which may not adequately support fastidious organisms or anaerobes. Although the sites were practising to HPA standards, there was considerable variation in collection and microbiology practice and this heterogeneity may be important. As part of this study, we attempted to ensure that all centres practised appropriate sample collection by developing and delivering to them a training package in person or remotely, and we updated staff when turnover occurred.

In almost two-thirds of patients (58%), there was a difference in the described biome from the microbiology culture results depending on whether swab or tissue sampling was used. Furthermore, in half of the patients (50.4%), there was a difference in the number of pathogens reported. However, this was not invariably that tissue samples had a greater yield (i.e. they reported the information contained in the paired swab sample) and additional organisms (although this was the case for 36.7% of patients). In a minority of cases (8.1%), the swab sample had a greater yield than the tissue sample, that is, it reported additional information over the tissue samples. This variation in yield may be attributable to variation in the bacterial profile across a wound surface. Therefore, a swab taken from an area of a tissue subsequently removed for tissue sampling would be a closer comparison, able to account of the spread of bacteria, although the area of tissue removal would preclude this practice in many wounds, as it would potentially impact healing.

A greater proportion of tissue sample results detailed at least one antibiotic to which pathogens within the sample were resistant (41.8% vs. 31.1%) and sensitive (67.8% vs. 55.9%); however, it was not always the case that the tissue samples reported all the information contained in the paired swab sample.

The fact that a tissue sample is able to collect bacteria from deep within the wound bed and that a swab relies on the capture of bacteria from the fluid expressed by pressing the wound (as per Levine et al. ’s48 technique) means that one might expect a tissue sample to collect additional deep bacteria. We found that tissue samples provide information on more pathogens, but it is not clear from this study if the added information might have an effect on clinical decision-making. Wound microbiology results are only one aspect of the clinical assessment, with direct assessment of the wound progress during treatment likely to be an important cue for determining progress or deterioration in a wound and guiding treatment. If clinicians currently default to swabbing wounds, then it is uncertain if a move to tissue sampling is warranted.

It is clear that more patients experienced a higher degree of pain with tissue sampling than swabbing, regardless of the ulcer type. In addition, post-sampling bleeding was noted more often after tissue sampling than swabbing – bleeding of concern was attributable to tissue sampling in 24 (6.0%) patients, attributable to swab sampling in 3 (0.8%) patients and attributable to both swab and tissue sampling in 3 (0.8%) patients. The limited information indicates that there is a small difference in costs for tissue sampling and swab sampling and, hence, the question remains as to the added clinical value of tissue sampling over swabbing.

One of the strengths of our study is that we recruited a large study population with a single aetiology, all clinically suspected to be infected and intended to be treated on antibiotics. As the question we sought to address (i.e. is the selection of sampling by swab or tissue collection best) arises in the management of infected foot ulcers, we have studied this group rather than a consecutive series of ulcers or a unselected sample. This is because bacterial sampling in chronic wounds is not used to diagnose infection; if it were, then a study sample that included both infected and uninfected wounds would be essential. One should not usually collect microbiology from a clinically uninfected wound and, therefore, only patients with clinically infected wounds were recruited. The study was pragmatic in that it allowed clinicians to diagnose infection according to their current clinical practice. This means that our study is relevant to contemporary practice in a range of settings and not just specialist centres. The processing of samples using current NHS laboratory practice is also a strength, as results are thus applicable to regular clinical settings.

Given the lack of a gold standard, our study did not consider diagnostic accuracy, but rather agreement; hence, we used appropriate statistical methods to summarise and analyse our data. Most previous studies asking a similar question have presumed that a tissue sample was the criterion standard against which swab specimens are judged as providing true- and false-positive results. We believe that this is not appropriate, as it presupposes that there is a criterion standard assumed to be the tissue sample, whereas there is evidence from studies that have found swabs to report additional isolates in 11% (Mutluoglu et al. 62) and 8.1% of samples (CODIFI), and different isolates in 6.7% (Mutluoglu et al. 62) and 13.2% (CODIFI).

Our study was also significantly larger than previous investigations in the area.

Overall, the results indicate that the results of tissue and swab cultures are different in a substantial minority of cases, with tissue sampling usually providing reports with higher numbers of pathogens. This is potentially attributable to the less detailed reporting by some microbiology laboratories for the swab specimens. These factors favouring tissue samples must be weighed against the slightly more complex process of collecting tissue specimens, the slightly greater pain and bleeding, and the possibly slightly higher cost.

Introduction

In the main study (see Chapter 2), the extent of agreement and pattern of disagreement on presence and number of pathogens reported was investigated, and we concluded that tissue samples more often reported additional information than swab samples. At the outset of the study, however, we were aware that if one method of sampling did provide more information as assessed by the number of pathogens identified, this did not necessarily mean that the additional microbiology information would be considered ‘more informative’ by clinicians. We therefore wished to investigate the clinician’s perspective on the microbiology reports. In order to determine whether or not a microbiology report from tissue sampling was ‘more informative’ than a swab report, we presented paired reports (a swab report and a tissue report from the same patient) to clinicians to assess the microbiology information within a patient vignette, and we analysed how the clinician would have responded to the available information. Therefore, this clinical panel review study investigated the clinical usefulness of the information provided by tissue versus swab samples, using an expert clinical panel blinded to the type of specimen to interpret the results. We assumed that a more informative culture report would allow a more appropriate decision regarding the choice of an empiric antibiotic regimen.

The best available test of the clinical utility of whether either swab or tissue information was better would be a trial in which clinicians were provided with results from one or the other sample, but this was not possible within the remit of this commissioned study.

If the method of sampling makes no difference to ongoing treatment decisions, then the sampling method that is the quickest, cheapest and best tolerated by patients would be the recommended method for clinical practice. However, if the choice of sampling method does affect subsequent treatment choice, then there is a trade-off between clinical usefulness and cost, specimen collection-related pain or bleeding and clinician time and skill in taking the sample.

Methods

Results

The clinical panel review involved a total of 13 reviewers and sample results from 250 patients.

Of the 250 patients whose microbiology results were included in the clinical review, 30 were also used as an inter-rater ‘validation’ sample to assess the reliability between reviewers, and another 30 were used as an intrarater ‘validation’ sample to assess reliability within reviewers (i.e. comparing their responses on the same data on two different occasions). This corresponded to a total of 310 paired patient vignettes reviewed (310 swab sample and 310 tissue samples). Three patients were excluded from the evaluable clinical review population: one from batch 1, as the patient was subsequently found to have a protocol deviation for which their swab sample was not sent to the laboratory; and two from batch 2, as a result of missing responses from reviewers. Therefore, a total of 247 patients were included in the evaluable clinical review population, corresponding to 307 paired patient vignettes across the main study and the inter- and intrarater reliability substudies. Figure 12 presents a diagram of the process, the number of patient vignettes and the number of reviewers involved in the clinical review. Patterns of missing data and samples of patient vignettes are presented in Appendix 2.

FIGURE 12.

Diagram of the clinical review process, population and reviewers.

Reviewers were each sent between 13 and 31 paired patient vignettes, provided in two batches, each consisting of two rounds (in order to separate vignettes from patients swab and tissue samples).

Table 33 presents a summary of all the reviewers involved in the clinical review and the number of patients whose vignettes (pairs of vignettes) were reviewed. This includes those forming the main clinical review sample used to address the comparison of the appropriateness of the empirical antibiotic regimen between swab and tissue samples and also the additional inter- and intrarater validation samples.

Association between baseline factors on extent of agreement on the requirement for a change/initiation in therapy

Multinomial regression modelling with a random effect for reviewers (and MI to allow for missing data) was used to assess whether extent of agreement, based on the overall summary of the requirement for a change or initial in therapy, was influenced by the pre-specified baseline covariates (Table 41).

TABLE 41 - Multinomial regression analyses for individually fitted baseline factors with random reviewer effect

AIC, Akaike information criterion.

Smaller is better.

Based on the reduction in –2 log-likelihood from the null model.

An odds ratio of > 1.0 indicates an increase in the odds of the specified outcome (change/initiation in therapy indicated by swab but not tissue, or by tissue but not swab) compared with the reference group (swab and tissue agree on change/initiation in therapy) for the specified factors.

Factors with missing data.

Baseline characteristics	‘Swab and tissue agree on requirement for change/initiation in therapy’ vs. a change/initiation in therapy indicated by	Odds ratioa (95% CI)	AICb	p-valuec
Null model			376.64
Patient on an empirical antimicrobial regimen: yes vs. no			379.68	0.6203
	1: Swab but not tissue	2.32 (0.29 to 18.48)
	2: Tissue but not swab	0.86 (0.30 to 2.50)
Ulcer type: any ischaemia (± neuropathy) vs. neuropathic only			379.68	0.6205
	1: Swab but not tissue	1.53 (0.62 to 3.77)
	2: Tissue but not swab	1.16 (0.59 to 2.25)
Ulcer grade			380.12	0.3408
Grade 2 vs. grade 1	1: Swab but not tissue	3.07 (0.92 to 10.22)
	2: Tissue but not swab	1.11 (0.48 to 2.54)
Grade 3/4/5 vs. grade 1	1: Swab but not tissue	2.17 (0.60 to 7.86)
	2: Tissue but not swab	1.43 (0.63 to 3.22)
Pre-sampling antibiotic therapy: yes vs. nod			380.47	0.9218
	1: Swab but not tissue	1.07 (0.43 to 2.65)
	2: Tissue but not swab	1.08 (0.54 to 2.15)
Antimicrobial dressing: yes vs. nod			379.97	0.7171
	1: Swab but not tissue	0.85 (0.34 to 2.15)
	2: Tissue but not swab	1.29 (0.62 to 2.71)
Wound duration: < 56 days vs. ≥ 56 daysd			379.69	0.6211
	1: Swab but not tissue	0.67 (0.27 to 1.69)
	2: Tissue but not swab	0.82 (0.41 to 1.64)
Log-wound duration (continuous)d			377.93	0.2576
	1: Swab but not tissue	1.23 (0.90 to 1.66)
	2: Tissue but not swab	1.15 (0.91 to 1.45)

None of the baseline factors was found to be significant at the 10% level. Therefore, there was no evidence of an association between any of baseline factors on the extent of agreement on the requirement for a change/initiation in therapy between swab and tissue samples.

The overall test of the covariance parameter based on the change in likelihood for the model with and without the random intercept for reviewer was not statistically significant (p-value = 0.5937). However, it remains important to account for the variation in the data attributable to the reviewer, and, therefore, the random effect for reviewer remained in the null model during testing of covariate effects. Figure 13 displays the ranked predicted random reviewer effect for the null model, for outcome ‘2 – Tissue but not Swab indicates change/initiation in therapy’ compared with the reference ‘3 – Swab and Tissue agree on change/initiation in therapy’.

FIGURE 13.

Plot of predicted random reviewer effect in the null model for outcome: tissue but not swab indicates change/initiation in therapy.

The figure presents the ranked predicted random reviewer effect for the null model. Note that the magnitude of the predicted reviewer effects are comparable to the natural log of the odds presented in Table 41. For example, for the tissue but not the swab indicating a change in therapy versus the swab and tissue agreeing on the requirement to change therapy for patients on an empirical antibiotic regimen against those who are not, the odds ratio is 0.86, with a log-odds of –0.15.

Discussion

This substudy set out to investigate the potential clinical impact of providing either a swab result or a tissue result to a clinician who was tasked with making a decision on antibiotic therapy for a patient with an infected DFU. The previous chapter discussed the finding that tissue samples usually identified more pathogens than swabs, but we wanted to determine if this equated to providing more clinically useful information. We therefore assessed if there was a difference in a swab sample compared with a tissue sample for providing information indicating that empiric antimicrobial therapy was adequate and if the reports from swab or tissues indicated whether or not a change in antimicrobial therapy was needed (including the need to initiate antibiotic). In order to understand the reliability of the data collected through the panel review, we also assessed the inter- and intrarater reliability for the clinical assessors making these judgements. We are not aware of any previous studies that have attempted to understand the potential clinical impact of differences in microbiology reports by using clinical vignettes in this way.

We found that one in five patients had swab results that indicated that all the pathogens were covered, but the tissue results indicated that all pathogens were not covered by the empirical antibiotic regimen. Correspondingly, clinicians reported that in nearly one in five patients (17.8%), the swab results indicated that no change in therapy was required, whereas the tissue results did suggest that a change in therapy was required. In a smaller number of cases (8.9%), the swab results indicated that a change in therapy was required when the tissue samples did not. These recommended changes to antibiotic regimen (including patients who had received no antibiotics at the outset) were based only on clinical vignette information and hence did not represent the full set of information that should be available to clinicians when reviewing the appropriateness of antibiotic therapy, such as change in ulcer state and changes in local and systemic findings of infection, for example.

Overall, we observed that 9% more tissue samples than swab samples indicated that a change in therapy was required (17.8% vs. 8.9%). Therefore, if the treating clinician had the tissue sample results rather than the swab results they would probably change the antibiotic regimen for 1 in every 11 patients. This is a potential ‘number needed to treat differently’ of 11. If the additional cost, skill and/or discomfort associated with tissue compared with swab sampling is perceived as modest, then for every 11 people in whom a tissue sample is taken (rather than a swab), a different treatment decision may be taken.

The inter-rater reliability results suggested slight to modest overall agreement between different reviewers (Krippendorff’s alpha of 0.35 and 0.66 for swab and tissue samples, respectively), indicating substantial variability between reviewers. This emphasises the importance, therefore, of ensuring that the same reviewer reviewed swab and tissue samples in the main clinical review.

There are a number of limitations to this work. As mentioned, we provided the reviewing clinicians with a limited data set which did not include all the information usually available in clinical practice to evaluate whether or not a change in antibiotic therapy was needed. We note, however, that the study also asked reviewing clinicians to make a judgement on whether or not empiric antibiotics were appropriate, and this judgement did not require additional clinical cues, such as ulcer status.

We provided blinded, paired information to clinicians and attempted to mask them to whether the result came from tissue or swab samples, to prevent bias. We do not, however, know if our blinding was successful. As tissue samples reported more pathogens, the reviewing clinicians might have guessed that longer reports (with more isolates) came from tissue samples and shorter reports from swab samples.

Given the increased attention on providing appropriate antibiotic therapy, these results may indicate a potentially higher rate of change in therapy (one would hope, but cannot guarantee, from broad-spectrum to targeted narrow-spectrum antimicrobials) following collection of a tissue result. It is not known, however, if a clinician would always initiate the change from broad- to narrow-spectrum antibiotics if an ulcer is improving, or what factors would prompt such a change (such as audit of prevalence of broad- and narrow-spectrum regimens). Identifying additional pathogens might lead to an increased used of broad-spectrum antimicrobials, thus threatening best practice antibiotic stewardship.

The current data also do not tell us anything about whether or not resampling after initiation of empiric antibiotics is needed to tailor the antibiotic regimen post receipt of the microbiology report. If a change/initiation in therapy is indicated or required, then should a patient’s ulcer be recultured when microbiology results are returned?

The fact that there were 9% of patients in whom swab results indicated a change/initiation in therapy when tissue results did not means that choosing either approach will potentially miss microbiology information that might lead to a change/initiation in therapy. Therefore, each method provides complementary information and, in certain circumstances, it might be appropriate to use both methods, for example if empiric therapy has not resulted in improvement.

It also needs to be borne in mind that tissue sampling may not be appropriate in all clinical cases or settings. In some patients, for example those with clotting disorders at high risk of bleeding, then the risk/benefit of tissue sampling over swabbing will differ from other patients.

Further work would be needed to determine whether clinical review of the swab and tissue results in these patients actually led to a real change in antibiotic regimen, rather than these ‘virtual panel’ results.

Introduction

Traditional methods of pathogen detection, such as microscopy of stained smears, plating specimens for culture and performing biochemical tests to identify the isolated organism, have major limitations. 86–89 These include the inability to culture some pathogens, a low sensitivity of pathogen detection,86–88 a lack of diagnostic specificity89 and the length of time and cost for processing. 89 Furthermore, in patients receiving antibiotic therapy, cultures may be falsely negative. Recent years have seen the advent of novel molecular (genotypic) methods, such as PCR, which have transformed the characterisation of organisms and diagnosis. 52 Traditional (phenotypic) culture methods may not identify minor, although possibly important, components of a mixed bacterial population; molecular methods provide an alternative with increased sensitivity (partly related to their ability to retrieve information on organisms that do not survive transport to the laboratory or are difficult to culture) and reduced time to identify the pathogens. 52,53,86,87,89

A study by Davies et al. 86 compared bacterial microflora of 18 (8 healing and 10 non-healing) chronic venous leg ulcers identified by using plating/culture methodologies and by PCR techniques. Although culture-based methods revealed that the majority of both wound types carried the aerobes Staphylococcus and Pseudomonas spp. (89% and 80% in healing and non-healing ulcers, respectively), PCR also identified strains not detected by plating/culture methods. The results of this86 study are consistent with a previous study87 in which the molecular approaches demonstrated significantly greater bacterial diversity of the wound microflora than that revealed by plating/culture methods. Thomsen et al. 90 also compared the plating/culture and PCR techniques in 14 chronic venous ulcers. They reported that PCR identified flora not found by traditional methods. Importantly, they found variation in the flora within the wound, such that a single sample may not represent the biome adequately. Similarly, other studies have consistently demonstrated the conflicting results of plating/culture methods and molecular techniques, the latter being more sensitive than the former. 91–93 More importantly, application of the molecular methods in guiding treatment enables a targeted approach to wound care, which has the potential to improve patient outcomes. 94–96

Ideally, wounds would be treated with antibiotics only after receiving results from microbial analysis, in order to limit overprescription and to ensure that narrow-spectrum antibiotics are used when possible. This is, however, only possible with rapid techniques, such as PCR. Furthermore, traditional culturing methods may be biased as a diagnostic tool as they select for easily cultured organisms, such as S. aureus, and select against difficult to culture bacteria, such as obligate anaerobes. 97 The main disadvantage of DNA-based PCR techniques is the inability to distinguish viable DNA sequences from inactive or dead organisms,98–100 unless supplementary methods are used. 101,102

This small substudy examined the identification of pathogens present in suspected infected DFUs using both conventional plating and culture-based techniques and PCR. This enabled us to investigate the agreement between the two analysis techniques (i.e. whether or not organisms identified by PCR reflected the bacterial load captured by plating/culture techniques). Moreover, we were able to obtain further detailed information on the agreement in identification of organisms between the two analysis techniques.

Methods

For patients included in the substudy, a second swab sample was taken and the tissue sample was cut in two so that samples obtained using either sampling technique could be analysed using both PCR and conventional plating/culture.

Results

Discussion

This study has found that PCR techniques, when used to analyse either swab or tissue samples, yield a higher number of reported pathogens than conventional plating and culture, with the number of pathogens reported by PCR dependent on the applied percentage cut-off of total microbial load within a sample, selected to be 5% in our study. Whether the higher number of pathogens is meaningful clinically is not clear, as the yield in PCR may be of persistent microbial DNA that is not associated with any current infection but that remains in the wound.

In our study, we used a single laboratory to analyse wound flora with PCR techniques. Although for this study the results were not available for clinical decision-making as they were batch processed, the results from such a laboratory can be returned within 24 hours. The arguments for greater use of PCR techniques (over culture and plating) are twofold: they make available increased amounts of information (often including identification of new bacterial species); and they allow for quicker analysis. This, however, needs to be set against the patchy availability and cost (which may change as technology advances) of such techniques. Furthermore, the impact of a higher yield from PCR on the complexity of laboratory results may affect clinical decision-making, particularly because clinicians have been accustomed to microbiology reports in which there has been selection of organisms for reporting (as identified in a small number of cases from our survey of practice) rather than reporting of the wound biome overall. The identification of a large number of species at low frequency provides clinical teams with increased data, this then requires selection of the meaningful information on likely pathogens associated with the current clinical picture, and hence ability to identify contaminants.

Although PCR techniques are often represented as a gold-standard, they too are vulnerable to contamination and (as per plating and culture) require standardised techniques and validated processes. There is a potential for false-positive results associated with the very high sensitivity. 103,104 Fenollar et al. 104 reported a comparison of plating and culture techniques in sampling for bone and joint infections (525 patients) and identified 50 discordant results: seven cases where different organisms were identified by the two analysis techniques, 21 where PCR only was positive (culture was negative), and 22 cases where the culture was positive (13 culture contaminants and nine ‘false-negative’ PCR results) and the PCR negative. They identified that in 475 out of 525 cases, the results were in agreement but their baseline event rate was much lower than in this study, as their patients did not have to be infected to enter the study. The utility of comparing PCR and plating/culture techniques in uninfected wounds is not clear given that these techniques are, currently, not used to determine the presence, or otherwise, of infection.

Introduction

Understanding the natural history of DFUs is an important pre-requisite to the planning of clinical trials to examine new interventions in the management of DFUs. It enables an informed approach when incorporating various design parameters, such as sample size estimates, follow-up period, frequency of visits and important outcomes.

The natural history of a condition can be reported from inception cohorts, but these are not commonly available in the DFU literature. The largest relevant prospective cohort study is the Eurodiale (European Study Group on Diabetes and the Lower Extremity) study,105 which recruited 1232 patients from 14 European countries. The study collected data on the characteristics of patients with DFUs, diagnostic and management procedures, health-care organisation, quality of life measurement and resource use. Patients with a new DFU were followed up monthly until healing of the foot, lower-leg amputation, death or non-healing after 1 year. 106 They reported that 58% of ulcers were infected at first presentation, with the prevalence of infection varying markedly between centres (range 28–74%); this may be due to referral patterns, standards of care or sensitivity of infection diagnosis. Data are reported on healing for 998 patients, of whom 77% healed within the follow-up period (12 months). This high rate of healing requires further confirmation, as the Eurodiale study reported higher healing and lower adverse outcome rates than earlier studies: only 5% of patients underwent a major amputation and 6% died. 106 This may reflect the potential reporting bias associated with a large proportion of missing data.

As well as overall prognosis, we also wanted to understand the relationship between patient and wound characteristics and outcomes. In the UK, Ince et al. 107 explored the relationships between time to healing of DFUs and baseline characteristics of the patients and their ulcers in a large prospective cohort study. They reported on 449 participants referred to a specialist clinic and found four variables independently associated with median time to healing: (1) area of the ulcer; (2) severity of peripheral arterial disease (PAD); (3) ulcer site; and (4) duration of diabetes.

In our previous systematic review,61 we identified that there were no data available specifically exploring the influence of infection status on the time to healing. Trials designed to investigate agents for treating DFUs, such as topical agents/antibiotics, may recruit populations that are homogeneous in respect of their infection status (infected ulcers or uninfected ulcers only). It would be useful to know the clinical outcomes in infected ulcers, as these are at highest risk of adverse sequelae. Characterising this patient population should allow both event rate estimation for trial planning, as well as providing data on prognosis that would be useful for the patients, their families and health-care service planning. It is essential that future trials commissioned in the area of DFUs are informed by accurate data on natural history, in the context of modern diabetes management protocols.

Some studies have reported that factors such as ulcer area, diabetes duration, severity of PAD and ulcer site provide prognostic information. However, the prognostic value of the newest tools, such as the PEDIS classification scheme, has not been assessed in this way. When Schaper et al. 108 examined the relationship between infection, neuropathy and ischaemia, they concluded that infection itself had no effect on healing in people with neuropathy, but it did have an effect on the healing of those with ischaemia (healing rate of 77% in people with ischaemia and no infection and 64% in those with ischaemia and infection). This suggests that treatment of infection in people with peripheral neuropathy may restore the potential for healing, whereas in people with lower extremity ischaemia, usual care for infection in the absence of addressing their ischaemia may be insufficient. This could be due to poor antibiotic penetration when administered systemically in those with PAD.

Schaper et al. 108 concluded that the profile of people with DFUs is changing, with a move away from the simple neuropathic foot to a more complex clinical presentation. This is in the context of an ageing population in which more people may have concurrent arterial disease. They note that more emphasis should be placed on developing strategies to improve outcome for neuroischaemic foot ulcers and, in particular, those with PAD and infection.

Methods

The CODIFI prognosis of foot infection follow-up study was an extension to the original multicentre cross-sectional CODIFI study design involving 400 patients with a DFU with suspected infection requiring antibiotic therapy64 (see Chapter 2).

The prognosis study involved the addition of a case-note review to allow identification of clinical outcomes 12 months after baseline swab and tissue sampling.

The prognostic substudy was submitted as a substantial amendment (version 4.0) for ethical approval on 19 October 2012. Approval was granted on 29 November 2012. By the time of this approval, 248 participants had been recruited into CODIFI. We therefore needed to obtain both prospective consent for subsequent participants for follow-up at 12 months (via a revised protocol sent for ethical review), as well as obtain retrospective consent for those recruited under the initial protocols (again via an application to the Research Ethics Committee).

Results

Medium-term outcomes

Cross-tabulations of patients’ healing status, against the other outcomes of death, amputation, revascularisation surgery and reoccurrence, are presented in Table 62. The index ulcer was reported to have healed in 136 (45.5%) patients; however, of these, 13 (9.6%) patients’ index ulcers then reoccurred within the follow-up period. A total of 45 (15.1%) patients died within the 12-month follow-up period, 52 (17.4%) had an amputation of the same limb on which the index ulcer was found (or on the same limb) and 18 (6.0%) had revascularisation surgery.

TABLE 62 - Cross-tabulation of healing against amputation, death, revascularisation surgery and reoccurrence within 12 months of sampling (these are not mutually exclusive outcomes)

N/A, not applicable.

	Index ulcer healed?		Total
	Yes	No
Patient died, n (%)
Yes	8 (2.7)	37 (12.4)	45 (15.1)
No	128 (42.8)	126 (42.1)	254 (84.9)
Total	136 (45.5)	163 (54.5)	299 (100.0)
Amputation (of or on the limb of the index ulcer), n (%)
Yes	12 (4.0)	40 (13.4)	52 (17.4)
No	124 (41.5)	123 (41.1)	247 (82.6)
Total	136 (45.5)	163 (54.5)	299 (100.0)
Revascularisation surgery, n (%)
Yes	8 (2.7)	10 (3.3)	18 (6.0)
No	128 (42.8)	153 (51.2)	281 (94.0)
Total	136 (45.5)	163 (54.5)	299 (100.0)
Index ulcer reoccurred, n (%)
Yes	13 (9.6)	N/A	13 (4.3)
No	123 (90.4)	N/A	123 (90.4)
Total	136 (100.0)	N/A	136 (100.0)

There were 12 (4.0%) patients whose index ulcer was reported to have healed and for whom an amputation of, or on, the same limb as the index ulcer was also reported; for two of these patients, the amputation occurred after the reported healing of the index ulcer, and for 10 patients the amputation occurred prior to reported healing of the index ulcer (Table 63). Although it may appear counterintuitive for there to be an amputation and later healing of the index ulcer, the area of amputation in relation to the index ulcer suggests a different amputation site to that of the index ulcer. Amputations were predominantly performed on one or more digits, confirming that healing was in reference to the index ulcer rather than the amputation site.

TABLE 63 - Summary of patients with index ulcer healed and amputation within 12 months post sampling

Summary	Total (N = 12), n (%)
Amputation after index ulcer healed without reoccurrence of index ulcer	2 (16.7)
Amputation before index ulcer healed	10 (83.3)

The time to each outcome by 12 months is presented in Table 64. Of the 136 patients whose index ulcer healed (excluding 12 patients whose date of healing was missing), the median time to healing was 4.5 months (range 0.5–12.9 months). Of the 12 patients whose ulcers were reported to have reoccurred after initially healing (excluding one patient with missing reoccurrence date), the median time to reoccurrence from healing was 1.7 months (range 0.3–10.7 months). The median time to death for the 45 patients was 5.6 months (range 0.6–11.5 months) and the median time to amputation of the index ulcer/limb for 52 patients was 2 months (range 0.0–10.6 months). Finally, of the 16 patients who had revascularisation surgery (excluding 2 patients whose date of surgery was missing), the median time to surgery was 3.0 months (range 0.1–9.5 months).

TABLE 64 - Time to outcome within 12 months post sampling (months)

SD, standard deviation.

Partial dates were imputed as the 15th of the month for nine patients’ dates of healing and three patients’ dates of amputation. Furthermore, two patients with inconsistent dates of reported healing were corrected by the Clinical Trials Research Unit based on other information provided for each patient.

Time to outcomea	Total
Time to first healing (n = 136)
Mean (SD)	5.5 (3.47)
Median (range)	4.5 (0.5–12.9)
Missing	12
Time from healing to reoccurrence (n = 13)
Mean (SD)	4.1 (4.01)
Median (range)	1.7 (0.3–10.7)
Missing	1
Time to death (n = 45)
Mean (SD)	5.9 (3.23)
Median (range)	5.6 (0.6–11.5)
Missing	0
Time to amputation (n = 52)
Mean (SD)	3.0 (3.05)
Median (range)	2.0 (0.0–10.6)
Missing	0
Time to revascularisation surgery (n = 18)
Mean (SD)	3.4 (2.85)
Median (range)	3.0 (0.1–9.5)
Missing	2

Table 65 presents the number and types of ‘other’ events as reported by the local team (podiatrists or research nurses). A total of 45 ‘other’ events were reported in 37 (12.4%) patients with substantial variability in the type of other events reported.

TABLE 65 - Summary of the number of patients with an ‘other’ event reported

Other events were reported for a further 13 patients; however, these events were not considered to be clinically relevant in the context of CODIFI. Such events included discharge to the community team, state of open wound, redressing of ulcer and hospital admissions with no detail of cause.

Other event occurred	Total (n = 299)
Yesa	37 (12.4%)
No	262 (87.6%)
Type of other event (non-mutually exclusive categories)
Charcot	2
Further amputation	2
Further healing following reoccurrence	2
Further ulcer(s)	11
Further ulcer(s) and healing	3
Incision and drainage	1
Kidney and pancreas transplant	1
Necrobiosis lipoidica to legs	1
Osteomyelitis	1
Skin graft and negative pressure therapy	1
Surgical correction of deformity	1
Debridement (surgical and non-surgical)	7
Contralateral amputation	4
Contralateral angioplasty	1
Contralateral gangrene	1
Contralateral healing of ulcer	1
Contralateral osteomyelitis	1
Contralateral revascularisation	1
Contralateral ulceration	3

A summary of patients’ first occurring outcome (healing, revascularisation, surgery, amputation, death) is presented in Table 66 and provides further detail of the outcomes reported subsequent to the first.

TABLE 66 - Summary of clinical outcome sequence within 12 months post samplinga

When healing is reported to have taken place following amputation, amputation is assumed to be of part of the index foot not containing the index ulcer.

Clinical outcome sequence	Total (N = 299), n (%)
Index ulcer healed	122 (40.8)
Healed only	110 (90.2)
Healed then revascularisation surgery	1 (0.8)
Healed and revascularisation (order unknown missing)	1 (0.8)
Healed then amputation	2 (1.6)
Healed then patient died	8 (6.6)
Revascularisation surgery	10 (3.3)
Revascularisation surgery only	1 (10.0)
Revascularisation surgery then healed	4 (40.0)
Revascularisation surgery then amputation	3 (30.0)
Revascularisation surgery then patient died	1 (10.0)
Revascularisation surgery then amputation then healed	1 (10.0)
Amputation (of or on the limb of the index ulcer)	46 (15.4)
Amputation only	26 (56.5)
Amputation then healed	8 (17.4)
Amputation then revascularisation surgery	4 (8.7)
Amputation then patient died	6 (13.0)
Amputation then revascularisation surgery then healed	1 (2.2)
Amputation and revascularisation surgery then patient died	1 (2.2)
Patient died (without prior events)	29 (9.7)
Healing, amputation or death not reported	92 (30.8)

Cumulative incidence of healing in the presence of competing risks: death and amputation

Table 67 presents the occurrence of amputation and death in patients whose index ulcer had not healed. Of the 163 (54.5%) patients whose index ulcer had not healed, 93 (57.1%) were known to be alive and without amputation at 12 months and were censored at the earliest of their case-note review or at 12 months post sampling; 33 (20.2%) were known to be alive with amputation before 12 months and seven (4.3%) were known to have died with amputation before 12 months. These patients therefore had a competing event at their date of amputation and 30 (18.4%) patients died without amputation before 12 months and therefore had a competing event at their date of death.

TABLE 67 - Healing within 12 months post sampling

Healing of the index ulcer	Total (N = 299)
Number of patients with index ulcer healed, n (%)
Yes	136 (45.5)
No	163 (54.5)
Index ulcer not healed, n (%)
Patient alive with amputation before 12 months	33 (20.2)
Patient alive without amputation at 12 months	93 (57.1)
Patient died with amputation before 12 months	7 (4.3)
Patient died without amputation before 12 months	30 (18.4)

Figure 14 shows the estimated cumulative incidence curves of the time to healing or the competing risks of death or amputation. Corresponding healing estimates at 6 and 12 months are presented in Table 68. The estimated 6 months’ post sampling healing rate was 27.5% (95% CI 22.4% to 32.5%) and the 12 months’ post sampling healing rate was 44.5% (95% CI 38.9% to 50.1%). The median time to healing for all patients in the follow-up population was not reached and is not estimated.

FIGURE 14.

Cumulative incidence functions of healing in the presence of competing risks: death and amputation.

TABLE 68 - Cumulative incidence of healing estimates at 6 and 12 months in the presence of competing risks: death and amputation

The number left at 12 months includes two participants known to have healed just after the 12-month time point, and an additional participant, owing to the variability in the multiply imputed data sets.

Month	Healing estimate (95% CI)	Number healed	Number with competing amputation event	Number with competing death event	Number censored	Number left
6	27.5% (22.4% to 32.5%)	82	32	17	0	168
12	44.5% (38.9% to 50.1%)	133	40	30	93	3a

A total of 43 (14.4%) patients had at least one missing baseline covariate or date of healing (details of missing data items and imputed healing times are provided in Appendix 3).