Diabetes Care Volume 37, October 2014 2693

Sue E. Gardner,1 Ambar Haleem,2 Cultures of Diabetic Foot Ulcers Ying-Ling Jao,1 Stephen L. Hillis,3,4,5 CARE/EDUCATION/NUTRITION/PSYCHOSOCIAL CLIN John E. Femino,6 Phinit Phisitkul,6 Without Clinical Signs of Infection Kristopher P. Heilmann,2 Shannon M. Lehman,1 and Do Not Predict Outcomes Carrie L. Franciscus5 Diabetes Care 2014;37:2693–2701 | DOI: 10.2337/dc14-0051

OBJECTIVE We examined associations between ulcer bioburden and ulcer outcomes in neu- ropathic diabetic foot ulcers (DFUs) that lacked clinical signs of infection.

RESEARCH DESIGN AND METHODS Three dimensions of bioburden (i.e., microbial load, microbial diversity, and the presence of likely pathogens) were measured at baseline using swab cultures obtained by Levine’s technique. Subjects were assessed every 2 weeks for 26 weeks to determine the rate of healing and development of infection-related complications. Foot ulcers were off-loaded using total-contact casts and routinely debrided. To establish associations between bioburden and rate of healing, Cox proportional hazards and least squares regression were used after adjusting for ulcer depth, surface area, and duration.

RESULTS A total of 77 subjects completed the study. Sixty-five (84.4%) had ulcers that healed during follow-up; weeks-to-closure ranged from 2 to 26 (median 4.0). Mean (6 SD) percent reduction in surface area/week was 25.0% (6 23.33). Five 1 (6.5%) of the DFUs developed an infection-related complication. None of the College of Nursing, The University of Iowa, Iowa City, IA bioburden dimensions (i.e., microbial load, microbial diversity, or presence of 2Department of Internal Medicine, The Univer- likely pathogens) was significantly associated with weeks-to-closure or percent sity of Iowa, Iowa City, IA reduction in surface area per week. Weeks-to-closure was best predicted by ulcer 3Department of Radiology, The University of duration, depth, and surface area (c-statistic = 0.75). Iowa, Iowa City, IA 4Department of Biostatistics, The University of CONCLUSIONS Iowa, Iowa City, IA 5Comprehensive Access and Delivery Research Culturing DFUs that showed no clinical signs of infection had no predictive value and Evaluation Center, Iowa City VA Health Care for outcomes of DFUs managed with total-contact casts and routine debridement. System, Iowa City, IA 6 These findings support recommendations of the Infectious Disease Society of Department of Orthopedic Surgery, The Univer- sity of Iowa, Iowa City, IA America that culturing and antibiotics should be avoided in treating DFUs that Corresponding author: Sue E. Gardner, show no clinical signs of infection. [email protected]. Received 7 January 2014 and accepted 10 June Early identification of diabetic foot ulcers (DFUs) destined for poor healing and/or 2014. infection-related complications remains problematic, often delaying and compromis- The views expressed in this article are those of ing treatment. Although current guidelines recommend antibiotic treatment be initi- the authors and do not necessarily reflect the ated when obvious clinical signs of infection develop (1), these signs may not appear position or policy of the Department of Veterans Affairs or the United States Government. until destruction of underlying tissue and bone triggers a systemic inflammatory re- © 2014 by the American Diabetes Association. sponse. Patients with diabetes, however, may not express clinical signs of infection, Readers may use this article as long as the work despite high levels of bacteria in local DFU tissue (1,2), because peripheral vascular is properly cited, the use is educational and not disease, poor metabolic control, and neuropathy dampen first-line inflammatory for profit, and the work is not altered. 2694 Bioburden and DFU Outcomes Diabetes Care Volume 37, October 2014

responses (3). Still, for patients with un- the development of DFU infection- antibiotics and enrolled the subjects 2 infected DFUs, indiscriminate use of anti- related complications (i.e., new wound weeks later. Those on long-term sys- biotics likely contributes to the growing deterioration, new osteomyelitis, or temic antibiotics for chronic non-DFU problem of antibiotic resistance (4). The- new amputation due to DFU infection). infections, such as chronic urinary tract oretically, in DFUs that show no clinical Three bioburden dimensions were ex- infections, were excluded from the signs of infection, the judicious and time- amined as predictors: 1)microbial study. Individuals meeting inclusion/ex- lier use of antimicrobial treatment might load; 2) microbial diversity; and 3) pres- clusion criteria were enrolled after pro- be guided by first assessing bioburden. In ence of S. aureus (including MRSA), ob- viding informed written consent. reality, however, it is unclear which di- ligate anaerobes, proteobacteria (i.e., Baseline data were collected immedi- mension(s) of bioburden are actually as- Gram-negative rods), and b-hemolytic ately after enrollment. sociated with poor healing or the Streptococcus. Ulcer dressings (i.e., Lyofoam; Molnlycke development of a DFU infection. Health Care), off-loading devices (i.e., Although bioburden has traditionally RESEARCH DESIGN AND METHODS total-contact casts, used for 72 subjects been used to refer to the number of Design and DH boots for 5 subjects), and de- microorganisms contaminating a sur- A prospective-cohort design was used. bridement (i.e., sharp debridement of face (5), wound bioburden is often Each dimension of ulcer bioburden was necrotic tissue in the wound bed and used to refer to three dimensions or as- measured at baseline, and the research callus on the wound edge) were standard- pects of the microbial community that team assessed the rate of healing and ized for all study subjects as a way to limit may contribute to poor healing and/or development of infection-related com- factors unrelated to ulcer bioburden and development of infection-related com- plications every 2 weeks until 1) the ul- minimize variability in DFU outcomes. plications (6). These dimensions include cer healed, 2)theDFUfootwas DFU management did not include anti- 1) microbial load (i.e., the number of amputated, 3) the subject was lost to microbial dressings, topical antimicro- organisms per gram of tissue); 2)micro- follow-up, or 4)after26weeksoffollow- bials, and/or systemic antibiotics, unless an bial diversity (i.e., the number of differ- up. The Institutional Review Board at infection-related complication occurred ent species); and 3) the presence of the University of Iowa approved study during follow-up. pathogenic organisms. Research sug- procedures. Subjects were enrolled gests, in DFUs and other types of chronic from September 2008 through February Study Variables wounds, a high microbial load (7,8) and 2012. Predictors: Three Dimensions of DFU microbe diversity (9) result in poor heal- Bioburden ing and infection. Common DFU patho- Setting and Sample Baseline measures of 1) microbial load, gens include Staphylococcus aureus Subjects were recruited through local 2) microbial diversity, and 3) presence of (10), methicillin-resistant S. aureus media advertisements (newspaper and likely pathogens were obtained to de- (MRSA) (11,12), Gram-negative bacilli television advertisements) and from termine ulcer bioburden. After cleans- (13), b-hemolytic Streptococcus (10), outpatient clinics at a large, academic- ing with nonbacteriostatic saline, ulcer ’ and obligate anaerobes (10); these mi- affiliated medical center and a Veterans specimens were collected using Levine s crobes are targeted for antibiotic treat- Affairs medical center. The research technique: An Amies swab (Copan, Bre- 2 ment in DFUs with moderate to severe team screened diabetic adults (i.e., scia, Italy) was rotated over a 1-cm area clinical signs of infection (1). Unfortu- $18 years of age) with a DFU on the of viable wound tissue for 5 s, using suf- fi nately, studies of DFU bioburden typi- plantar surface of the foot and excluded cient pressure to extract wound-tissue fl cally have focused on only one or two any subjects meeting the following ex- uid, and immediately transported in a dimensions of bioburden; however, to clusion criteria: 1)significant ischemia (i.e., charcoal transport to a research micro- determine which, if any, of the dimen- toe-brachial index or ankle-brachial biology laboratory. The swab was vor- sions of bioburden are important in pre- pressure index ,0.5); 2) signs or symp- texed in 1 mL of trypic soy broth, and dicting poor healing or the development toms of clinical DFU infection (i.e., in- the resulting suspension was processed of infection-related complications, it is creasing pain, erythema, heat, edema, to measure three dimensions of bio- necessary to examine all three dimen- or purulent exudate) or osteomyelitis burden using the procedures described sions of bioburden in a single cohort of (i.e., positive radiograph); and 3) treat- below. subjects. ment with systemic antibiotics in the Microbial Load Accordingly, the purpose of this study prior 2 weeks. Presence of osteomyelitis Each suspension was serially diluted in was to determine whether the three di- was assessed using X-rays of the ulcer tryptic soy broth and each dilution mensions of wound bioburden can be foot at the time of screening. Excluding plated on Columbia blood agar (Remel, used to predict outcomes among per- individuals on systemic antibiotics min- Lenexa, KS), eosin-methylene blue agar sons who have neuropathic DFUs but imized any influence of antibiotics on (EMB; Remel), and CHROMagar MRSA no clinical signs of DFU infection. Our baseline ulcer bioburden. Some individ- (Becton Dickinson, Sparks, MD) plates. specific aims were to examine the asso- uals met all inclusion and none of the Columbia and EMB plates were incu- ciations between baseline measures of exclusion criteria, except having been bated in 5% CO2 at 378C for 48 h, and each dimension of bioburden (listed be- administered systemic antibiotics in MRSA plates were incubated aerobically low) and 1) DFU rate of healing (i.e., the 2 weeks prior despite the lack of at 378C for 48 h. Each dilution was also weeks-to-closure and percent reduction clinical signs of DFU infection. In these plated onto Brucella Agar supplemented of ulcer surface area per week) and 2) cases, the research team discontinued with blood, hemin, and vitamin K (Remel) care.diabetesjournals.org Gardner and Associates 2695

and incubated in an anaerobe jar at 378C wound,” a valid and reliable definition white count, elevated erythrocyte sedi- for 48 h. The species of infecting organ- (17). Agreement between the assess- mentation rate, or elevated C-reactive isms was identified via standard micro- ments was 98.7%. The principal investi- protein at follow-up visits. If these indi- biological procedures (14). Because gator resolved any discrepancies between cators were absent at follow-up, radio- dilutions are based on a single swab, the raters. graphs were not retaken. Subjects the plate count of each species was mul- To compute changes in ulcer size, ul- experiencing new amputations had tiplied by the dilution factor to yield total cers were measured at baseline and, if their medical records reviewed by the number of colony-forming units (CFUs) the ulcer was not healed, at each follow- research team (J.E.F. and P.P.) to ensure for that species. Microbial load was de- up visit. Two members of the research amputations were due to DFU infection fined as the sum CFUs of all species or team independently assessed size using and not some other reason. total CFUs per swab. the VeVMD digital software system (Vista Medical, Winnipeg, Manitoba, Demographic and Secondary Variables Microbial Diversity Canada) and procedures previously de- At baseline, the research team collected Microbial diversity was defined as the scribed (18). Inter- and intrarater reli- demographic variables (age, sex, and number of different species identified ability of VeVMD was 0.76 and 0.87 race), smoking history (packs per day from both aerobic and anaerobic plates. (Pearson r), respectively (18). A cotton- and years of smoking), diabetes type tipped swab, placed in the deepest as- Presence of Likely Pathogens and duration, and duration of the study fi pect of the DFU, was marked where the S. aureus was identi ed on Columbia ulcer using subject self-report and med- swab intersected with the plane of the blood agar, based on the appearance ical records. Standard laboratory tests b peri-wound skin. The distance between of characteristic yellow -hemolytic col- were used to measured baseline glyce- thetipoftheswabandthemarkwas onies, which appeared as Gram-positive mic control (HbA ) and nutritional sta- measured as ulcer depth using a centi- 1c cocci organized into grapelike clusters tus (albumin and prealbumin levels). At meter ruler. on stain and tested catalase positive as baseline, trained members of the re- well as Staphylococcus latex-agglutination search team assessed neuropathy (5.07 positive. To identify MRSA strains, all Development of Infection-Related Semmes-Weinstein monofilament). Mi- Complications S. aureus isolates were screened by crocirculation (transcutaneous oxygen Development of infection-related com- PCR for the mecA gene, according to pressure) was measured at baseline plications was defined as wound deteri- previously published methods (15,16). and at each follow-up visit using a trans- oration, new osteomyelitis, and/or a Organisms that grew anaerobically on cutaneous oxygen monitor (Novametrix new amputation due to DFU infection. supplemented Brucella agar, but not 840; Novametrix Medical Systems Inc.). aerobically, were identified as anaer- Subjects whose wounds deteriorated obes. Organisms that grew on EMB and/or developed new osteomyelitis Data Analysis plates and stained Gram-negative were treated for infection and moni- Rate of Healing were identified as proteobacteria tored until either their ulcer healed, they DFU rate of healing was defined using (Vitek Legacy; Biomerieux, Durham, were lost to follow-up, or at the end of the two metrics: 1)weekstocomplete NC). b-hemolytic, catalase-negative, 26-week follow-up period. Study partici- wound closure and 2) percent reduction Gram-positive cocci were classified to pation ended after amputation. in ulcer surface area per week. The as- Lancefield group (A, B, C, F, and G) using Wound deterioration was defined as sociation between each dimension of the PathoDX Strep Grouping Kit (refer- the new development of erythema and bioburden (i.e., microbial load, micro- ence 62076; Remel). frank heat and an increase in size .50% bial diversity, and presence of each over baseline. Ulcer size was measured pathogen) and the number of weeks to Outcomes: DFU Rate of Healing and as described above. Two members of complete wound closure were exam- Development of Infection-Related the research team independently as- ined using Cox proportional hazards re- Complications sessed each DFU for erythema and frank gression, with weeks-to-closure treated Outcomes were measured every 2 heat, and agreement between raters as censored when subjects were lost to weeks during follow-up. The members was 92.5 and 95.5%, respectively. The follow-up before healing. We regressed of the research team who assessed principal investigator resolved any dis- weeks-to-closure separately on each di- the rate of healing and development of crepancies. Members of the research mension of bioburden. Both unadjusted infection-related complications were blind team assessed other clinical signs of in- and adjusted models were fit. For ad- to DFU bioburden status at all follow-up fection, including increasing pain, justed models, baseline measures of ul- visits. edema, and purulent drainage. These cer duration, depth, and surface area Rate of Healing signs and symptoms were not used to were used as covariates because they Therateofhealingwasdefined as: 1) define wound deterioration because are associated with rate-of-healing out- weeks-to-closure (complete healing); they only occurred in one subject who comes (19). To assess the additional use- and 2) percent reduction in ulcer surface also had erythema and heat. Develop- fulnessprovidedbyeachbioburden area per week. Two of several members ment of new osteomyelitis was assessed dimension, we also fit a model contain- of the research team independently as- using radiographs and magnetic reso- ing only covariates (i.e., ulcer duration, sessed ulcer closure at each study visit nance imaging when subjects presented depth, and surface area) as indepen- using the Wound Healing Society’s with new tracts to bone, wound deteri- dent variables. We described effects of definition of “an acceptably healed oration, elevated temperature, elevated each dimension of bioburden with 2696 Bioburden and DFU Outcomes Diabetes Care Volume 37, October 2014

relative risk (RR). We described the dis- described the bioburden associated (15.6%) on the midfoot, and 5 (6.5%) crimination ability of the proportional with ulcers that developed an infection- on the heel. hazards models using the c-statistic, as related complication. Table 2 summarizes baseline measures discussed by Harrell et al. (20,21). The c- for each dimension of bioburden. Six statistic estimates the probability that the RESULTS (7.8%) DFU specimens produced no model correctly discriminates between Recruitment and Enrollment growth following incubation. These two subjects having different weeks to Ninety-six individuals met the inclusion ulcers had microbial load and diversity val- wound closure. criteria and were screened for eligibility. ues of 0 and none of the likely pathogens. The association between bioburden Twelve of these were subsequently ex- Subject follow-up ranged from 2 to 26 and percent reduction in surface area cluded due to osteomyelitis (n = 6), long- weeks, with a mean of 7.2 (SD 6 6.3) per week was determined as in the term antibiotics for chronic infections weeks. During follow-up, 65 (84.4%) ul- weeks-to-wound-closure analyses, ex- (e.g., chronic urinary tract infection; cers healed, 1 (1.3%) resulted in an am- cept that a least squares regression n = 3), ischemia (toe-brachial index or putation, 8 (10.4%) were unhealed when was used. Adjusted models included ankle-brachial pressure index #0.5; n = lost to follow-up, and 3 (3.9%) were un- the covariates described above. We de- 1), clinical signs of DFU infection (n =1), healed at the end of 26 weeks. DFU out- scribed effects of each dimension of bio- and inability to use the off-loading de- comes are summarized in Table 2. burden with regression coefficients. We vice (n = 1). The remaining 84 persons described the discrimination ability of were enrolled in the study, including 17 Bioburden and Rate of Healing 2 the model with R . (20.0%) who were initially on systemic Results from analyses of the association These analyses were considered ex- antibiotics for a clinically uninfected between each dimension of bioburden ploratory, so we did not control for DFU. These 17 subjects were enrolled and weeks-to-closure are displayed in typeIerror.Allanalyseswereper- 2 weeks after discontinuing antibiotics. Table 3. Unadjusted results are in sec- formed using SAS 9.3 (SAS Institute Of the 84 enrolled subjects, 7 were ex- tion 1, adjusted results in section 2, and Inc., Cary, NC) using PROC LOGISTIC cluded from analyses due to missing results for the model with only covari- and PROC PHREG. Because PROC PHREG baseline or follow-up data. Seventy- ates in section 3. does not compute c-statistic for Cox pro- seven subjects were included in final Microbial load and microbial diversity fi portional hazards regression, it was analyses. were not signi cant predictors of the computed using added SAS statements. number of weeks to wound closure. Sim- An a = 0.05 was used for all analyses. Descriptive Statistics of Sample ilarly, presence of S. aureus, MRSA, an- Table 1 contains descriptive statistics for aerobes, proteobacteria, or b-hemolytic Development of Infection-Related study subjects and ulcers. The mean age Streptococcus did not predict the Complications of subjects was 54.9 (SD 6 11.6) years, number of weeks to wound closure. In- Only five subjects developed infection- and they were predominantly white (n = terestingly, when growth of any organ- related complications, making the 70; 90.9%), male (n = 62; 80.5%), and ism versus no growth was examined, power too low for testing associations type 2 diabetic (n = 66; 85.7%). All the presence of any organism was sig- between each dimension of bioburden (100%) subjects had neuropathy by nificantly associated with more weeks- and the development of infection- monofilament testing. Sixty (77.9%) ul- to-closure in unadjusted analyses (RR related complications. Alternatively, we cers were located on the forefoot, 12 0.34; P =0.02;c-statistic = 0.54) and

Table 1—Subject and ulcer characteristics at baseline (N = 77) Subject characteristics Normal values Mean (6 SD)/median (range) Age (years) 54.9 (6 11.64) Smoking (pack-years; n =75) 16.7 (6 26.15) Duration of diabetes (years; n =76) 15.9 (6 11.87) WBC (mm3)3,700–1,050 7,898 (6 1,834.00)

HbA1c (%; mmol/mol) 4.8–6.0 8.2 (6 1.95); 69 (6 5.00) Albumin (g/dL) 3.4–4.8 4.1 (6 0.30) Prealbumin (mg/dL; n = 75) 18–45 21.3 (6 4.85) C-reactive protein (mg/dL) ,0.5 2.4 (6 5.37) Erythrocyte sedimentation rate (mm/h; n = 76) Males: 0–15 31.1 (6 22.93) Females: 0–20 Ulcer characteristics Ulcer area (cm2) 1.9 (6 2.58)/1.2 (0.03–16.7) Ulcer depth (cm) 0.3 (6 0.28)/0.2 (0–1.0) Ulcer duration prior to study participation (weeks) 33.3 (6 40.90)/19.5 (0.5–156.0) Wound tissue oxygen pressure (mmHg; n = 75) 47.1 (6 14.38) Toe-brachial index (n =53) 0.8 (6 0.2) Ankle-brachial index (n =26) 1.0 (6 0.2) care.diabetesjournals.org Gardner and Associates 2697

Table 2—Baseline ulcer bioburden and ulcer outcomes after follow-up (N = 77) Mean (6 SD) or n (%) Median (range)

Bioburden dimensions Microbial load (total CFU/swab) 1.0 3 106 (6 3.75 3 106)5.13 104 (0–2.7 3 107) Microbial diversity (number of different species/swab) 3.6 (6 2.37) 3.0 (0–9.0) Potential pathogens Number (%) of ulcers with S. aureus 28 (36.4) Number (%) of ulcers with MRSA 8 (10.4) Number (%) of ulcers with proteobacteria 27 (35.1) Number (%) of ulcers with b-hemolytic Streptococcus 19 (24.7) Number (%) of ulcers with anaerobes 15 (19.5) Ulcer outcomes Rate of healing Ulcers achieving complete closure/healing [n (%)] 65 (84.4) Weeks to wound closure [mean (6 SD)/median (range)] (n = 65) 6.0 (6 4.81)/4.0 (1.9–25.9) Percent reduction in surface area/week [mean (6 SD)/ median (range)] 25.0 (6 19.46)/23.3 (214.5 to 53.9) Developed infection-related complication [n (%)] 5 (6.5) Wound deterioration [n (%)] 4 (5.2) Osteomyelitis [n (%)] 0 (0.0) Amputation [n (%)] 1 (1.3) Six of the 77 subjects had no growth on culture plates. Therefore, the microbial load and microbial diversity for these subjects was 0. The mean and median for microbial load and microbial diversity were computed for the entire sample, including those with no growth. Therefore, the range includes 0 as the lower level. nearly significantly associated with the covariate-only model (R2 = 0.27), ul- complete closure. However, Ince et al. more weeks-to-closure in the adjusted cer depth and surface area had signifi- (19) included some nonplantar DFUs be- analyses (RR 0.42; P = 0.06; c-statistic = cant (P = 0.001 and 0.004, respectively) low the malleolus in their sample, which 0.74). Nevertheless, the model with negative associations (regression coeffi- may explain the longer time-to-heal re- only covariates showed slightly better cient = 224.29 and 22.33, respectively) ported in their study. discrimination in predicting weeks-to- with percent reduction in surface area Similarly, in our study, none of the closure (c-statistic = 0.75) than the ad- per week. Ulcer duration had a close to dimensions of bioburden predicted the justed model containing any organism significant (P = 0.05) and slightly positive percent reduction in ulcer surface area versus no growth and all three covari- association with percent reduction in ul- per week. In contrast, Xu et al. (8) found ates. In the covariate-only model, ulcer cer surface area per week (regression high microbial load was inversely re- depth and surface area had a significant coefficient 0.10). lated to the percent change in ulcer (P = 0.02 and 0.006, respectively) posi- area per day in a sample of neuropathic tive association with more weeks to Bioburden and Development of DFUs; however, their analyses did not wound closure (RR 0.30 and 0.81, re- Infection-Related Complications control for ulcer duration, depth, and spectively). Ulcer duration had a close Five (6.5%) subjects developed an infection- surface area. In our study, we found ul- to significant (P = 0.05) but slightly neg- related complication. The type of com- cer duration, depth, and surface area did ative association with more weeks to plication and bioburden data for these indeed predict weeks-to-closure and wound closure (RR 1.01). subjects are provided in Table 5. percent reduction in surface area per Results from analyses of the association week. In fact, the predictive power of between each dimension of bioburden and CONCLUSIONS these three variables was substantial in percent reduction in surface area per week None of the three dimensions of biobur- modeling the number of weeks to DFU are displayed in Table 4. Unadjusted results den (i.e., microbial load, microbial diver- closure (c-statistic = 0.75). In addition, are presented in section 1, adjusted results sity, and presence of potential pathogens) although Xu et al. (8) reported that sub- in section 2, and results for the model that predicted the number of weeks to ulcer jects in their sample were provided with includes only covariates in section 3. closure (before or after adjusting for ulcer regular care, including debridement, it is None of the dimensions of bioburden duration, depth, and surface area). Sixty- unclear if any, or which, off-loading (i.e., microbial load, microbial diversity, five (84.4%) of DFUs in this study healed techniques were used. All subjects in our or presence of potential pathogens) was during the 6-month follow-up period, study were off-loaded using a total-con- significant in predicting percent reduc- 50% of which healed in #4 weeks. This tact cast, and DFUs were sharp-debrided tion in surface area per week in either is considerably shorter than the 9 weeks on a regular basis. Total-contact casting the unadjusted or adjusted analyses. (63 days) reported by Ince et al. (19), even (22) and routine debridement (23) likely The three covariates had significant or though the DFUs in their retrospective contributed to the high number of subjects close to significant associations with study were similar in terms of inclusion/ who healed, the rapid rates of healing, percent reduction in surface area per exclusion criteria, baseline ulcer size, ulcer and the low number of infection-related week in all of the adjusted models. In management, and proportion achieving complications observed in this study. 2698 Bioburden and DFU Outcomes Diabetes Care Volume 37, October 2014

Table 3—Proportional hazards regression results for weeks to wound closure (N = such as Pseudomonas aeruginosa,are 77) believed to be common pathogens of Covariate P values chronic wounds (10), and indeed, we frequently found that species in our P value RR LCL UCL c-Stat Duration Depth Area study subjects’ infected ulcers. Of the Unadjusted models 15 ulcers that harbored anaerobes, the Predictor condition of 1 deteriorated during Microbial load 0.23 1 1 1 0.51 Microbial diversity 0.42 0.96 0.86 1.06 0.54 the study. Finally, of the 19 DFUs har- Presence of any 0.02 0.34 0.14 0.81 0.54 boring b-hemolytic Streptococcus,only S. aureus 0.31 0.77 0.46 1.28 0.53 2 developed an infection-related com- MRSA 0.94 0.97 0.44 2.14 0.51 plication despite widespread concern Anaerobes 0.81 0.92 0.48 1.78 0.49 that b-hemolytic Streptococcus is a Proteobacteria 0.12 0.66 0.39 1.12 0.54 common pathogen in DFUs (25). To- b-hemolytic Streptococcus 0.74 1.10 0.62 1.94 0.49 gether, these findings indicate that Adjusted models DFUs harboring the pathogens for which Predictor Microbial load 0.53 1.00 1.00 1.00 0.75 0.077 0.036 0.009 weassayedneednotbetreatedwith Microbial diversity 0.39 0.95 0.85 1.06 0.74 0.053 0.036 0.008 antibiotics unless the abscess displays Presence of any 0.06 0.42 0.17 1.03 0.74 0.056 0.021 0.012 clinical signs of a pathogenic infection. S. aureus 0.18 0.70 0.41 1.18 0.74 0.038 0.013 0.008 An antibiotic-free approach is particu- MRSA 0.29 0.64 0.28 1.46 0.74 0.030 0.014 0.007 larly recommended when rigorous off- Anaerobes 0.97 1.01 0.51 2.02 0.75 0.057 0.025 0.009 loading and debridement are part of Proteobacteria 0.19 0.70 0.41 1.2 0.73 0.049 0.038 0.008 the wound-care regimen. b-hemolytic Streptococcus 0.94 1.02 0.57 1.85 0.73 0.044 0.025 0.011 This is the first study to prospectively Model with no predictor, all three covariates examine whether any of the three di- Covariate mensions of bioburden predict out- Ulcer duration 0.05 1.01 1.00 1.01 0.75 comes among neuropathic DFUs that Ulcer depth 0.02 0.30 0.11 0.83 show no clinical signs of infection. Our Ulcer surface area 0.009 0.81 0.69 0.95 data suggest that bioburden does not Each line represents one model. Microbial load is the total number of CFUs per swab. Microbial correlate with DFU outcomes. Nonische- diversity is the number of different species per swab. Presence of any is a dichotomous indicator mic, neuropathic ulcers represent 70% b for growth vs. no growth cultures. S. aureus, MRSA, anaerobes, proteobacteria, and -hemolytic (27) of all DFUs and therefore are most Streptococcus are indicators for presence of the corresponding organism. A significant covariate P value indicates that the covariate is useful in the model. The c-statistic in the model with no frequently encountered by clinicians. predictor (all three covariates) is for the whole model. RR or hazard function with RR ,1 Typically, persons with DFUs present to indicates that presence of an organism or larger values are associated with more weeks-to- their care provider for treatment when closure. c-Stat, c-statistic; LCL, lower 95% confidence limit for RR; P value, P value for testing H0: RR = 1; UCL, upper 95% confidence limit for RR. they are free from clinical infection, and the findings of this study suggest that culturing their wounds to assess biobur- In the absence of these management swab, a wider range than the often-cited den (including microbial load) is unnec- strategies, a high microbial bioburden threshold of 106 organisms/g of tissue essary, further supporting the Infectious may have a greater impact on the rate (25) believed to be indicative of infec- Diseases Society of America (IDSA) of DFU healing. tion. Microbial diversity also varied, guideline that DFUs without clinical signs To our knowledge, this is the first ranging from one to nine microbial spe- of infection should not be treated with study to examine wound deterioration cies. Of the nine species detected in one antibiotics (1). Nevertheless, before our as an outcome in a prospective DFU DFU, four were common pathogens, in- subjects were enrolled in our study, 20% study. Among the five (6.5%) DFUs that cluding b-hemolytic Streptococcus. De- were being treated with antibiotics with developed an infection-related compli- spite treatment with antibiotics, that no clear clinical rationale. Such practices cation, the dimensions of bioburden DFU remained unhealed at the end of likely contribute to the growing problem varied greatly. Four wounds deterio- the 6-month follow-up period. of antibiotic resistance (4). The findings rated, while only one (1.3%) resulted in In our study, of the 28 ulcers that har- of this study stress the need to translate an amputation, a rate much better than bored S. aureus, including MRSA, only 4 the IDSA guideline into practice. the 4.9% amputation rate seen among developed an infection-related compli- Many published classification sys- Medicare beneficiaries with both neuro- cation. (The DFU that harbored MRSA tems are used for diagnosing DFU infec- pathic and ischemic foot ulcers (24). Our resulted in an amputation.) A previous tions. All are based on clinical symptoms lower rate is likely due to an absence of study demonstrated that species iso- and signs of inflammation, the extent of significant macro- and microvasular lated from ischemic DFUs were likely the ulcer, and the host response to the compromise, as well as our routine prac- to contain antibiotic-resistant S. epidermi- inflammation. Of the variety of classifi- tice of off-loading and debridement. dis (90%) (26). Of 27 study ulcers that har- cation systems, the IDSA system is eas- Among five DFUs that developed an bored proteobacteria (Gram-negative iest to use and has been prospectively infection-related complication, microbial bacteria), the condition of only three validated as predicting the need for hos- load ranged from 104 to 107 microbes/ deteriorated. Gram-negative bacteria, pitalization (28–31). care.diabetesjournals.org Gardner and Associates 2699

Table 4—Least squares regression results for percent reduction in surface area per week (N = 77) Covariate P values P value Coefficient LCL UCL RSQR Duration Depth Area Unadjusted models Predictor Microbial load 0.24 0.00 0.00 0.00 0.02 Microbial diversity 0.32 20.95 22.83 0.93 0.01 Presence of any 0.08 214.47 230.73 1.79 0.04 S. aureus 0.32 24.61 213.79 4.58 0.01 MRSA 0.47 25.33 219.86 9.20 0.01 Anaerobes 0.68 2.32 28.90 13.53 0.00 Proteobacteria 0.27 25.20 214.44 4.05 0.02 b-hemolytic Streptococcus 0.96 20.26 210.52 10.00 0.00 Adjusted models Predictor Microbial load 0.79 0.00 0.00 0.00 0.27 0.052 0.002 0.004 Microbial diversity 0.67 20.37 22.07 1.34 0.27 0.052 0.003 0.004 Presence of any 0.25 28.53 223.18 6.12 0.28 0.066 0.002 0.005 S. aureus 0.42 23.28 211.37 4.81 0.27 0.055 0.002 0.004 MRSA 0.50 24.35 217.18 8.48 0.27 0.041 0.002 0.004 Anaerobes 0.48 3.51 26.40 13.43 0.27 0.043 0.001 0.005 Proteobacteria 0.94 0.33 28.19 8.85 0.27 0.048 0.002 0.004 b-hemolytic Streptococcus 0.96 20.21 9.43 9.02 0.26 0.037 0.003 0.005 Model with no predictor, all three covariates Covariate Ulcer duration 0.05 0.10 0.00 0.20 0.27 Ulcer depth 0.001 224.29 238.92 29.65 Ulcer surface area 0.004 22.33 23.86 20.79 Each line represents one model. Microbial load is the total number of CFUs per swab. Microbial diversity is the number of different species per swab. Presence of any is a dichotomous indicator for growth vs. no growth cultures. S. aureus, MRSA, anaerobes, proteobacteria, and b-hemolytic Streptococcus are indicators for presence of the corresponding organism. A significant covariate P value indicates that the covariate is useful in the model. The R2 in the model with no predictor (all three covariates) is for the whole model. Coefficient, estimated regression coefficient with a negative value indicating that presence or larger values are associated with smaller percent reduction in surface area per week; LCL, lower 95% 2 confidence limit for the regression coefficient; P value, for testing H0: regression coefficient = 0; RSQR, R ; UCL, upper 95% confidence limit for the regression coefficient.

A potential limitation of this study is that contributed to ulcer outcomes. of this study are relevant in clinical set- our culture-based method for measur- Nevertheless, our results are clinically tings that predominantly use culture- ing bioburden. We realize that standard relevant because other methods for based techniques. culture methods are limited in delineat- measuring bioburden (e.g., genomic A second limitation of this study ing true bioburden (32), and our culture- techniques and molecular assays, such stems from controversy of how best to based techniques might fail to detect as PCR) are not yet widely available in take a sample that assesses bioburden. microbial species and communities the clinical setting. Therefore, the findings One guideline suggests that tissue

Table 5—Bioburden of DFUs developing infection-related complications (N =5) Identification Type of Microbial load (total Microbial diversity (number of Potential number complication CFU/swab) species/swab) pathogens End of study reason 132 Wound deterioration 1.1 3 105 7 MRSA Lost to follow-up, unhealed Proteobacteria b-hemolytic Streptococcus 141 Wound deterioration 1.7 3 107 9 S. aureus Unhealed Anaerobes Proteobacteria b-hemolytic Streptococcus 176 Wound deterioration 6.0 3 105 1 S. aureus Unhealed 304 Wound deterioration 4.6 3 104 2 Proteobacteria Lost to follow-up, unhealed 165 Amputation 5.6 3 105 7 MRSA Amputation 2700 Bioburden and DFU Outcomes Diabetes Care Volume 37, October 2014

specimens must be obtained (1), while bioburden and DFU outcomes should diabetic foot ulcers. Clin Microbiol Infect 2006; another suggests that validated swab therefore use longitudinal designs. 12:186–189 specimens can be used (25). Cultures of 13. Aragon-S´ anchez´ J, Lipsky BA, Lazaro-´ Mart´ınez JL. Gram-negative diabetic foot oste- tissue are impractical in many clinics be- Funding. The National Institute of Nursing omyelitis: risk factors and clinical presentation. cause obtaining viable wound tissue suit- Research at the National Institutes of Health Int J Low Extrem Wounds 2013;12:63–68 able for culture is too invasive. Therefore, (R01-NR-009448 to S.E.G) supported this study. 14. Versalovic J. Manual of Clinical Microbiol- 54% of clinicians obtain swab specimens Duality of Interest. No potential conflicts of ogy. Washington, DC, ASM Press, 2011 when assessing wound bioburden (33). interest relevant to this article were reported. 15. Mendes RE, Kiyota KA, Monteiro J, et al. Author Contributions. S.E.G. designed the Rapid detection and identification of metallo- We previously found swab cultures ob- study, supervised study protocols, assisted beta-lactamase-encoding genes by multiplex tained using Levine’stechniquedatech- with data analyses, and wrote the manuscript. real-time PCR assay and melt curve analysis. J nique that expresses wound fluid from A.H. wrote parts of the manuscript. Y.-L.J. Clin Microbiol 2007;45:544–547 deep tissue layersdaccurately mea- collected research data, wrote parts of the 16. Richter SS, Heilmann KP, Dohrn CL, et al. sures all three dimensions of bioburden manuscript, and conducted parts of data anal- Activity of ceftaroline and epidemiologic trends ysis. S.L.H. conducted data analyses and wrote (i.e., microbial load, microbial diversity, in Staphylococcus aureus isolates collected parts of the manuscript. J.E.F. and P.P. enrolled from 43 medical centers in the United States and presence of pathogens) when com- subjects and managed subject care. K.P.H. in 2009. Antimicrob Agents Chemother 2011; pared with cultures of wound tissue completed all microbiological analyses. S.M.L. 55:4154–4160 (34). Specifically, the accuracy of swab coordinated study protocols and edited the 17. Margolis DJ, Berlin JA, Strom BL. Inter- cultures for measuring microbial load manuscript. C.L.F. managed study data and observer agreement, sensitivity, and specificity edited the manuscript. S.E.G. is the guarantor “ ” was0.80comparedwithbiopsy;micro- of a healed chronic wound. Wound Repair of this work and, as such, had full access to all Regen 1996;4:335–338 bial diversity based on swabs was 3.0 the data in the study and takes responsibility for 18. Gardner SE, Frantz RA, Hillis SL, Blodgett TJ, species/wound compared with 3.1 spe- the integrity of the data and the accuracy of the Femino LM, Lehman SM. Volume measures cies/wound based on biopsy cultures; data analysis. using a digital image analysis system are reli- and the concordance in identifying po- able in diabetic foot ulcers. Wounds 2012;24: References tential pathogens was 96% for S. aureus, 146–151 b 1. Lipsky BA, Berendt AR, Cornia PB, et al.; In- 19. Ince P, Game FL, Jeffcoate WJ. Rate of heal- 99% for -hemolytic Streptococcus,and fectious Diseases Society of America. 2012 In- ing of neuropathic ulcers of the foot in diabetes 96% for P. aeruginosa. fectious Diseases Society of America clinical and its relationship to ulcer duration and ulcer A third limitation is that we did not practice guideline for the diagnosis and treat- area. Diabetes Care 2007;30:660–663 consider how DFU outcomes might be ment of diabetic foot infections. Clin Infect Dis 20. Harrell FE Jr, Califf RM, Pryor DB, Lee KL, – influenced by biofilms. Biofilms are poly- 2012;54:e132 e173 Rosati RA. Evaluating the yield of medical tests. 2. Gardner SE, Hillis SL, Frantz RA. Clinical – microbial communities enclosed in a JAMA 1982;247:2543 2546 signs of infection in diabetic foot ulcers with 21. Harrell FE Jr, Lee KL, Mark DB. Multivariable polysaccharide matrix that is secreted high microbial load. Biol Res Nurs 2009;11: prognostic models: issues in developing models, by the bacteria. This environment permits 119–128 evaluating assumptions and adequacy, and bacteria-to-bacteria signaling (i.e., quo- 3. Apelqvist J. Diagnostics and treatment of the measuring and reducing errors. Stat Med – rum sensing) and a synergistic regulation diabetic foot. Endocrine 2012;41:384 397 1996;15:361–387 4. Howell-Jones RS, Price PE, Howard AJ, 22. Katz IA, Harlan A, Miranda-Palma B, et al. A of virulence factors that renders the bac- Thomas DW. Antibiotic prescribing for chronic fi randomized trial of two irremovable off-loading teria in bio lms more resistant to host skin wounds in primary care. Wound Repair Re- devices in the management of plantar neuro- – defenses and antimicrobial treatment gen 2006;14:387 393 pathic diabetic foot ulcers. Diabetes Care (35). Increasingly, biofilms are believed 5. Salcido R. What is bioburden? The link to 2005;28:555–559 chronic wounds. Adv Skin Wound Care 2007; 23. Lebrun E, Tomic-Canic M, Kirsner RS. The to play an important role in wound chro- 20:368 role of surgical debridement in healing of dia- nicity. Measuring biofilms, however, is 6. Gardner SE, Frantz RA. Wound bioburden betic foot ulcers. Wound Repair Regen 2010;18: very complex and requires epifluores- and infection-related complications in diabetic 433–438 foot ulcers. Biol Res Nurs 2008;10:44–53 cence microscopy and assays of quorum- 24. Margolis DJ, Malay DS, Hoffstad OJ, et al. 7. Browne AC, Vearncombe M, Sibbald RG. sensing signal molecules (35), techniques Prevalence of diabetes, diabetic foot ulcer, and High bacterial load in asymptomatic diabetic lower extremity amputation among Medicare not widely available in clinical settings. patients with neurotrophic ulcers retards fi Moreover, prospective studies have yet wound healing after application of Dermagraft. bene ciaries, 2006 to 2008. Diabetic foot ulcers. to define the relationship between biofilms Ostomy Wound Manage 2001;47:44–49 In Data Points #1 (prepared by the University of 8. Xu L, McLennan SV, Lo L, et al. Bacterial load Pennsylvania DEcIDE Center, under Contract No. and DFU outcomes (35). HHSA29020050041I). Rockville, MD, Agency for Finally, we did not measure wound predicts healing rate in neuropathic diabetic foot ulcers. Diabetes Care 2007;30:378–380 Healthcare Research and Quality, February 2011 bioburden at times of ulcer deteriora- 9. Trengove NJ, Stacey MC, McGechie DF, Mata [AHRQ Publication No. 10(11)-EHC009-EF] tion. This might be problematic because S. Qualitative bacteriology and leg ulcer healing. 25. Steed DL, Attinger C, Colaizzi T, et al. Guide- the microbial milieu of a chronic wound J Wound Care 1996;5:277–280 lines for the treatment of diabetic ulcers. – typically evolves over time, in response 10. Bowler PG, Duerden BI, Armstrong DG. Wound Repair Regen 2006;14:680 692 26. Galkowska H, Podbielska A, Olszewski WL, to environmental pressure. Depending Wound microbiology and associated ap- proaches to wound management. Clin Micro- et al. Epidemiology and prevalence of methicillin- on the virulence of new inhabitants biol Rev 2001;14:244–269 resistant Staphylococcus aureus and Staph- and their interactions within the micro- 11. Eleftheriadou I, Tentolouris N, Argiana V, ylococcus epidermidis in patients with diabetic bial community, the wound may fail to Jude E, Boulton AJ. Methicillin-resistant Staph- foot ulcers: focus on the differences between heal altogether. Therefore, measuring ylococcus aureus in diabetic foot infections. species isolated from individuals with ischemic Drugs 2010;70:1785–1797 vs. neuropathic foot ulcers. Diabetes Res Clin bioburden over time may provide better 12. Tentolouris N, Petrikkos G, Vallianou N, Pract 2009;84:187–193 insight than a single baseline assessment. et al. Prevalence of methicillin-resistant Staph- 27. Reiber GE, Vileikyte L, Boyko EJ, et al. Causal Future studies of the association between ylococcus aureus in infected and uninfected pathways for incident lower-extremity ulcers in care.diabetesjournals.org Gardner and Associates 2701

patients with diabetes from two settings. Dia- diabetic foot ulcer classification system for re- 33. Bamberg R, Sullivan PK, Conner-Kerr T. Di- betes Care 1999;22:157–162 search purposes to predict lower extremity am- agnosis of wound infections: Current culturing 28. Johnson SW, Drew RH, May DB. How long to putation. Int J Diabetes Dev Ctries 2009;29:1–5 practices of U.S. wound care professionals. treat with antibiotics following amputation in 31. Lavery LA, Armstrong DG, Murdoch DP, Wounds 2002;14:314–328 patients with diabetic foot infections? Are the Peters EJ, Lipsky BA. Validation of the Infectious 34. Gardner SE, Frantz RA, Saltzman CL, Hillis SL, 2012 IDSA DFI guidelines reasonable? J Clin Diseases Society of America’s diabetic foot in- Park H, Scherubel M. Diagnostic validity of three Pharm Ther 2013;38:85–88 fection classification system. Clin Infect Dis swab techniques for identifying chronic wound 29. Wukich DK, Hobizal KB, Brooks MM. Sever- 2007;44:562–565 infection. Wound Repair Regen 2006;14:548–557 ity of diabetic foot infection and rate of limb 32. Gardner SE, Hillis SL, Heilmann K, Segre JA, 35. Han A, Zenilman JM, Melendez JH, et al. The salvage. Foot Ankle Int 2013;34:351–358 Grice EA. The neuropathic diabetic foot ulcer importance of a multifaceted approach to char- 30. Widatalla AH, Mahadi SE, Shawer MA, microbiome is associated with clinical factors. acterizing the microbial flora of chronic wounds. Elsayem HA, Ahmed ME. Implementation of Diabetes 2013;62:923–930 Wound Repair Regen 2011;19:532–541