Predictors of Monomicrobial Necrotizing Soft Tissue Infections

Abstract

Background: Broad-spectrum antibiotic therapy is critical in the management of necrotizing soft tissue infections (NSTI) in the emergency setting. Clindamycin often is included empirically to cover monomicrobial gram-positive pathogens but probably is of little value for polymicrobial infections and is associated with significant side effects, including the induction of Clostridium difficile colitis. However, there have been no studies predicting monomicrobial infections prior to obtaining cultures. The purpose of this study was to identify independent predictors of monomicrobial NSTI where the use of clindamycin would be most beneficial. We hypothesized that monomicrobial infections are characterized by involvement of the upper extremities and fewer co-morbid diseases.

Methods: We reviewed all cases of potential NSTI occurring between 1996 and 2013 in a single tertiary-care center. The infection was diagnosed by the finding of rapidly progressing necrotic fascia during debridement with positive cultures of tissue. Univariable analysis was performed using the Student t-, Wilcoxon rank sum, χ², and Fisher exact tests as appropriate. Multivariable logistic regression was used to identify independent variables associated with outcomes.

Results: A group of 151 patients with confirmed NSTI with complete data was used. Of the monomicrobial infections, 61.8% were caused by Group A streptococci, 20.1% by Staphylococcus aureus, and 12.7% by Escherichia coli. Of the polymicrobial infections, E. coli was involved 13.7% of the time, followed by Candida spp. at 12.9%, and Bacteroides fragilis at 11.3%. On univariable analysis, immunosuppression, upper extremity infection, and elevated serum sodium concentration were associated with monomicrobial infection, whereas morbid obesity and a perineal infection site were associated with polymicrobial infection. On multivariable analysis, the strongest predictor of monomicrobial infection was immunosuppression (odds ratio [OR] 7.0; 95% confidence interval [CI] 2.2–22.3) followed by initial serum sodium concentration (OR 1.1; 95% CI 1.0–1.2). Morbid obesity (OR 0.1; 95% CI 0.0–0.5) and perineal infection (OR 0.3; 95% CI 0.1–0.8) were independently associated with polymicrobial infection.

Conclusion: We identified independent risk factors that may be helpful in differentiating monomicrobial from polymicrobial NSTI. We suggest empiric clindamycin coverage be limited to patients who are immunosuppressed, have an elevated serum sodium concentration, or have upper extremity involvement and be avoided in obese patients or those with perineal disease.

Necrotizing soft tissue infection (NSTI) is a rapidly progressive condition primarily involving subcutaneous fat, muscle, and fascia [1]. This disease was first described by Hippocrates in the Fifth Century before the Common Era and was first reported in the United States in 1871 by Confederate Army surgeon Dr. Joseph Jones, who described “hospital gangrene” with a mortality rate of 50% [2]. This uncommon condition (500–1,500 cases per year) continues to have a high mortality rate, 25%–35%, and is the subject of significant public and media interest [3,4].

Necrotizing soft tissue infection is categorized into three types on the basis of microbiologic culture data. Type I infections (polymicrobial) make up approximately 80% of NSTIs, whereas Type II (monomicrobial; classically caused by Streptococcus pyogenes) and Type III (clostridial infections) make up 15% and <5%, respectively [5]. Although useful in guiding antibiotic coverage, these data are not available to assist with empiric antibiotic choices [5–11].

Several studies have evaluated the risk factors for NSTI, which include age, diabetes mellitus, immunosuppression, peripheral vascular disease, obesity, and malnutrition. However, there have been no studies that evaluate risk factors for Type II (monomicrobial Group A Streptococcus [GAS] or Staphylococcus aureus infection) vs. Type I (polymicrobial) infections [12–20]. Because GAS represents greater than 50% of cases of monomicrobial NSTI, known presenting symptoms and anatomic/physiological markers would enable better empirical antibiotic choices, such as the inclusion or not of clindamycin. The current standard of care prescribes early detection, early implementation of empiric broad-spectrum antibiotics (often including clindamycin, a penicillin, vancomycin, and linezolid or daptomycin), and early, aggressive surgical debridement [6].

The purpose of this study was to identify predictors of monomicrobial NSTI in order to facilitate rapid and focused antibiotic therapy, specifically targeting GAS. Additionally, by identifying GAS NSTI infection on admission, we aim to limit unnecessary antibiotic use in potentially polymicrobial NSTI. Focused antibiotic therapy is important because it not only decreases overall cost but limits unnecessary exposure to broad-spectrum antibiotics, therefore decreasing the risk of adverse effects of therapy, specifically Clostridium difficile overgrowth.

Patients and Methods

With approval from our Institutional Review Board, we identified and reviewed all cases of potential NSTI occurring between 1996 and 2013 in a single tertiary-care center. Our search was limited to International Classification of Disease-9 codes (728.86, 40, 608.83). Only confirmed cases with complete data were analyzed. We defined true NSTI as a rapidly progressing infection demonstrating necrotic fascia with “dishwater” purulence at the time of debridement coupled with positive cultures of tissue. Patients were grouped according to the presence of either a monomicrobial or a polymicrobial infection. Patient age, race, co-morbidities, site of infection, physiologic data on admission, and the use of clindamycin or hyperbaric oxygen as part of the treatment regimen were recorded. Our primary outcome was the presence of monomicrobial infection. The secondary outcomes were mortality rate, length of stay (LOS), and the number of operations required.

Univariable analysis was performed using the Student t-, Wilcoxon rank sum, χ², and Fisher exact tests as appropriate. Multivariable logistic regression was used to identify independent variables associated with the presence of a monomicrobial infection. Predictor variables were identified using automated backward selection analysis.

Results

A series of 151 patients with confirmed NSTI with complete data was evaluated. Of these infections, 30.7% were monomicrobial. Of the monomicrobial infections, 61.8% were caused by GAS, 20.1% by S. aureus, and 12.7% by Escherichia coli. Of the polymicrobial infections, E. coli was involved 13.7% of the time, followed by Candida spp. at 12.9%, and Bacteroides fragilis at 11.3%. Of note, GAS also was involved 11.3% of the time in polymicrobial infections. On univariable analysis, immunosuppression, upper extremity infection, and initial serum sodium concentration >135 mmol/L were associated with monomicrobial infection, whereas morbid obesity and perineal involvement were associated with polymicrobial infection (Table 1).

Table 1.

Univariable Analysis (Number [%] of Patients Unless Otherwise Specified)^a

	Polymicrobial (n=86)	Monomicrobial (n=65)	p
Demographics
Mean age±SEM	52.5±1.35	51.0±1.86	0.49
Female	30 (35.0)	19 (29.2)	0.46
Caucasian	66 (76.7)	53 (81.5)	0.48
African American	18 (20.9)	11 (16.9)	0.54
Co-Morbidities
Current smoker	32 (37.2)	29 (44.6)	0.36
Alcohol abuse	24 (27.9)	24 (36.9)	0.24
Intravenous drug abuse	5 (5.8)	2 (3.1)	0.70
Diabetes mellitus type I or II	47 (54.7)	30 (46.2)	0.30
Immunosuppressed	7 (8.1)	19 (29.2)	0.0007
All obesity	45 (52.3)	15 (23.1)	0.0003
Obese	23 (26.7)	14 (21.5)	0.46
Morbidly obese	22 (25.6)	1 (1.5)	<0.0001
Human immunodeficiency virus infection	0	1 (1.5)	0.43
Neurologic disease	11 (12.8)	4 (6.2)	0.27
Cardiac disease	23 (6.7)	13 (20.0)	0.34
Hypertension	25 (29.1)	15 (23.1)	0.41
Hyperlipidemia	23 (26.7)	9 (13.8)	0.05
Peripheral vascular disease	5 (5.8)	3 (4.6)	1
Pulmonary disease	11 (12.8)	7 (10.8)	0.60
Hepatic disease	6 (7.0)	2 (3.1)	0.47
Hematologic disease	7 (8.1)	2 (3.1)	0.30
Renal failure	10 (11.6)	12 (18.5)	0.24
Active malignant disease	7 (8.1)	3 (4.6)	0.52
Thyroid disease	9 (10.5)	3 (4.6)	0.19
Transplant history	0	2 (3.1)	0.18
Hyperbaric oxygen treatment	8 (9.3)	4 (6.2)	0.56
Clindamycin use	54 (62.8)	51 (78.5)	0.04
Recent/active transfusion	30 (34.9)	23 (35.4)	0.95
Malnourished	0	0	–
Sites
Perineum	41 (47.7)	13 (20.0)	0.0004
Lower extremity	19 (22.1)	23 (35.4)	0.07
Upper extremity	8 (9.3)	17 (26.2)	0.006
Trunk	21 (24.4)	10 (15.4)	0.17
Head/neck	5 (5.8)	6 (9.2)	0.42
Admission Physiologic Data
Pressor requirement	10 (11.6)	13 (20.0)	0.16
Temperature (°C)(IQR)	37.4 (36.6, 38.5)	37.4 (36.8, 38.5)	0.43
Heart rate±SEM	98.7±1.9	104.4±2.7	0.09
Systolic blood pressure (mmHg) (IQR)	118 (102, 133)	115.5 (101, 131)	0.80
White blood cell count (×103/mcL)(IQR)	16.8 (21.3, 23.2)	14.6 (9.3, 21.0)	0.14
Hemoglobin (g/dL)±SEM	11.4±0.2	11.8±0.2	0.26
Serum sodium (mEq/L)±SEM	133±0.5	134.9±0.6	0.02
Serum creatinine (mg/dL) (IQR)	1.1 (0.9, 1.8)	1.2 (1.0, 2.4)	0.15
Serum glucose (mg/dL) (IQR)	138 (103, 229)	120 (97, 189)	0.12

^aParametric data are listed as mean±standard error of mean (SEM). Non-parametric data are listed as median (interquartile range [IQR]).

Patients with polymicrobial infections had a significantly lower mortality rate than did patients with a monomicrobial infection (9.3% vs. 21.5%) (Table 2). We noted no differences in the rate of subsequent C. difficile infection or the number of debridements required between monomicrobial and polymicrobial infections or in LOS based on infection type (see Table 1). Interestingly, patients with monomicrobial infections received clindamycin 78.5% of the time, whereas patients with polymicrobial infections received clindamycin 62.8% of the time (Table 1).

Table 2.

Outcomes by Type of Infection^a

	Polymicrobial (n=86)	Monomicrobial (n=65)	p 151
Death (%)	8 (9.3)	14 (21.5)	0.03
Clostridium difficile infection (%)	3 (3.5)	4 (6.2)	0.46
No. of operations	2 (1, 3)	2 (1, 3)	0.70
Length of stay	14 (10, 29)	15 (8, 25)	0.84

^aParametric data are listed as number (%). Non-parametric data are listed as median (interquartile range).

Multivariable analysis demonstrated that active immunosuppression and initial serum sodium concentration were independently associated with the presence of a monomicrobial infection (Table 3). Morbid obesity and involvement of the perineum were associated with polymicrobial infection.

Table 3.

Multivariable Analysis^a

	Wald χ2	Odds Ratio (95% Confidence Interval)	p
Immunosuppressed	11.2	7.3 (2.3–23.4)	0.001
Morbidly obese	7.0	0.1 (0.1–0.5)	0.007
Perineal infection	8.0	0.3 (0.1–0.7)	0.01
Initial serum Na concentration	6.0	1.1 (1.1–1.2)	0.008

^aC-statistic 0.82.

Discussion

Classic symptoms of NSTI are significant localized pain, local erythema, and swelling [6]. However, only 10%–40% of patients present with these classic symptoms [6]. Therefore, it is important to have a low threshold for early, aggressive antibiotic and operative treatment, because necrotic skin changes usually occur once skin ischemia and underlying necrosis have advanced well into the later stages of the disease [6,21]. Current literature shows multiple risk factors for NSTI, including, but not limited to, diabetes mellitus, alcohol use, intravenous drug abuse, immunosuppression, and peripheral vascular disease [22–24]. To our knowledge, no risk factors have been identified that are specific for Type II NSTI at the time of presentation. Expeditious and directed treatment of suspected organisms may limit the initial cost of treatment by avoiding the unnecessary use of broad-spectrum antibiotics and clindamycin, which may result in serious side effects.

Given the rapid progression of monomicrobial NSTI, it is imperative to begin antibiotic therapy as soon as possible. The standard of care currently consists of empiric broad-spectrum antibiotics effective against a wide variety of organisms. Traditionally, at our institution, patients with suspected NSTI are started on empiric piperacillin-tazobactam, vancomycin, and clindamycin. In suspected monomicrobial NSTI, isoxazole penicillin should be used, especially if it is not possible to rule out S. aureus involvement. In our study, 12.7% of monomicrobial necrotizing infections yielded S. aureus. If GAS or S. aureus Type II NSTI is suspected, it is imperative to start a protein synthesis-inhibiting antibiotic, most commonly clindamycin. Our study showed that GAS was involved in 61.8% of the monomicrobial infections, so patients who present with upper extremity infection, compromised immune status, and an elevated serum sodium concentration probably should be started on clindamycin in addition to a penicillin.

Clindamycin is a licosamide antibacterial drug that inhibits ribosomal translocation by acting at the 50S rRNA of the large ribosome subunit [25]. It inhibits both M protein and exotoxin production, thus facilitating phagocytosis, and suppresses the synthesis of tumor necrosis factor α (TNF-α), resulting in quelling of an excessive immune response [26]. The Eagle effect describes the effective limitations of penicillin after a streptococcal infection has reached a steady state. Clindamycin is not affected by the size of the bacterial inoculum or the stage of growth and thus has superior tissue penetration [27–32]. Clindamycin is more effective than penicillin against streptococcal species, and one study showed that the use of clindamycin was valuable in NSTI [33,34]. Additionally, GAS NSTI has a high risk of concomitant streptococcal toxic shock syndrome (STSS), which can lead to whole body cytokine dysregulation (cytokine storm), increasing the mortality risk [10]. It is the ability of clindamycin to suppress this cytokine response and toxin production that makes it such a powerful antibiotic against GAS [35].

Although clindamycin shows a clear advantage in the management of Type II NSTI, overuse in suspected Type I infections can cause considerable side effects and additional cost. One study found that 67% of surgical ward patients in whom C. difficile colitis developed had received clindamycin [36]. Kofsky et al. calculated that each episode of C. difficile colitis added an additional $2,000 to $5,000 to the cost of healthcare [37]. The goal of this study was to identify variables that could help diagnose Type II NSTI early in its course, thereby preventing the unnecessary use of clindamycin in patients with Type I infections.

Similar studies have shown a prevalence of S. pyogenes in monomicrobial infections. The study by McHenry et al., also showed S. pyogenes was isolated in only 17% of all NSTIs but accounted for 53% of the monomicrobial infections [10]. This is similar to our study, where S. pyogenes was isolated in 61.5% of our Type II NSTI cultures. Their study also showed that infection with S. pyogenes had a lower mortality rate than the overall mortality rate (18% vs. 29%, respectively), which differs from the mortality rate of 21.5% caused by Type II infections in our study. This difference could be explained by the fact that their mortality rate of 18% is dependent on S. pyogenes alone vs. all other forms of NSTI, whereas our study compared the mortality rate of Type II infections (21.5%) with that of Type I infections (9.3%). We cannot specifically attribute the deaths solely to S. pyogenes, even though they accounted for 61.5% of our Type II infections.

Since the publication by McHenry et al. in 1995, there has been a steady increase in the virulence of S. pyogenes, as shown by Stevens et al. [24]. Interestingly, other prior studies have shown that there is no difference in the mortality rates of gram-positive pathogens (9.09%) and gram-negative pathogens (17.14%; p=1.00) in monomicrobial infections, suggesting that S. pyogenes is no more virulent in necrotizing infections than are organisms such as E. coli or B. fragilis [38].

Wall et al. found a negative predictive value of 99% that a white blood cell count (WBC) <15,400/mm³ and a serum sodium concentration >135 mmol/L on admission ruled out NSTI with 90% sensitivity [39]. Wong et al. compared laboratory parameters and created a scoring system that could distinguish NSTI accurately from superficial cellulitis (Table 4) [40]. Although our study involves only patients with NSTI, we did find that an elevated serum sodium concentration had a positive predictive value for Type II NSTI.

Table 4.

The Laboratory Risk Indicator for Necrotizing Soft Tissue Infection

Variable	Score
C-reactive protein (mg/dL)
<15	0
≥15	4
Total white blood cell count (103 per mm3)
<15	0
15–25	1
>25	2
Hemoglobin (g/dL)
>13.5	0
11–13.5	1
<11	2
Sodium (mmol/L)
>135	0
≤135	2
Creatinine (mg/dL)
≤1.59	0
>1.59	2
Glucose (mg/dL)
≤180	0
>180	1

The Laboratory Risk Indicator for Necrotizing Soft Tissue Infection Adapted from reference 43. Values for each of the six parameters on the left (which may require unit conversion at different institutions) are assigned a score shown on the right. A total score of <6 suggests a 96% negative predictive value for deep necrotizing soft tissue infection (NSTI), and a score of ≥6 suggests a 92% positive predictive value for such infection.

There are several limitations of this study and its application to clinical practice. The percentage of patients who will present with all three indications of a monomicrobial infection (upper extremity, immunosuppression, and elevated serum sodium concentration) is small. Although, according to our data, there is a greater incidence of GAS in monomicrobial infections, there still is a 11.3% chance that a polymicrobial infection will involve GAS. At our institution, 78.5% of monomicrobial infections were managed with clindamycin compared with 62.8% of polymicrobial infections, indicating a trend toward limiting the use of clindamycin in suspected polymicrobial infections.

In conclusion, we believe that patients who present upper extremity infections, immunosuppression, and an elevated serum sodium concentration are ideal candidates for the addition of clindamycin or other protein synthesis-inhibiting antimicrobial agents, whereas patients who presented with suspected polymicrobial infections can forgo clindamycin.

Funding and Author Disclosure Statement

This work was funded by National Institutes of Health grant T32 AI078875. None of the authors has any conflict of interest with regard to this manuscript.