Cellulitis, clinical management, pathophysiology, review
Cellulitis (Latin: cellula (diminutive of cella: cell) + itis (suffix denoting inflammation)) and its subtype erysipelas (Greek: erythrós (red) + pella (skin)), are among the most frequent infections requiring hospitalisation.1 The historical distinction between cellulitis and erysipelas, based on different bacterial aetiologies and thus treatment options, is becoming obsolete as increasing evidence suggests a large overlap between these two entities (textbox 1). In the Netherlands, the annual incidence is estimated to be 22 per 1000 inhabitants. Approximately 7% of all patients with cellulitis are hospitalised.5,6 The mortality rate of hospitalised patients has been reported to be around 2.5%.7 Recent epidemiology data on cellulitis in the Netherlands are lacking, but given the rise in the incidence of important risk factors (namely diabetes, obesity and old age), an increase in the incidence of cellulitis is expected.8-10 Dutch guidelines on the clinical management of cellulitis of the lower extremities have been available since 2013 (figure 1).11 Since their publication, numerous studies have provided novel insights and new antibiotics registered for skin and soft tissue infections have entered the market. This review discusses the current state of evidence regarding pathogenesis, diagnostics, and treatment of cellulitis. The literature search strategy used is documented in textbox 2.
Cellulitis: a diagnostic challenge
All that is red is not cellulitis. The classical symptoms of erythema, oedema, warmth and tenderness, are non-specific and vary in severity. The clinical presentation of cellulitis is mimicked by a whole range of diseases (table 1 and figure 2). One recent study revealed that 31% of patients hospitalised with cellulitis were misdiagnosed, the most frequent mimickers being stasis dermatitis, stasis ulcers, gout, congestive heart failure, non-specific oedema and deep venous thrombosis (DVT).18 Another study in the primary care setting found a similar rate of misdiagnoses.19 Furthermore, when clinicians specifically consulted dermatologists because of uncertainty about a diagnosis of cellulitis, 74% of the patients turned out not to have cellulitis.20 Misdiagnosis results in unnecessary admissions and extra costs for perceived refractory cellulitis.18 Leucocytosis and elevated C-reactive protein (CRP) levels are present in 34-50% and 77-97% of patients, respectively.21,22
Stasis dermatitis can mimic all symptoms, including mild leucocytosis and/or CRP elevation. Its origin lies in chronic venous insufficiency, which causes proliferation and increased permeability of dermal capillaries. Leucocytes migrate, cause inflammation, stimulate collagen production, and thus induce dermal fibrosis.23 Erythrocyte extravasation causes brown skin pigmentation.24 Untreated, stasis dermatitis can progress to lipodermatosclerosis, which is characterised by a fibrotic tightening and sometimes ulceration of the skin above the ankles. Compression therapy can correct haemodynamic effects and cytokine levels.13
Imaging is sometimes indicated. As DVT does not occur more often in patients with cellulitis than those without, routine DVT screening is not recommended.25 When ultrasound was only utilised among uncertain ‘DVT vs cellulitis’ diagnoses, 17% turned out to have DVT.26 Ultrasound may detect occult abscesses, or disprove ‘abscesses’ mistakenly diagnosed during physical examination.27 Computed tomography is not warranted due to nonspecific findings.28 Magnetic resonance imaging and the Laboratory Risk Indicator for Necrotising Fasciitis score might help distinguish between necrotising fasciitis and cellulitis, but as yet neither have proven superior to clinical suspicion and subsequent surgical exploration.29 Uncomplicated superficial abscesses with erythema can be difficult to distinguish from primary cellulitis with secondary abscesses.30 Uncomplicated abscesses are treated with incision and drainage,31 but two recent trials show cure rate increases from 69-74% to 81-83% with adjunctive antibiotics.32,33
Multiple physical barriers and active protective mechanisms prevent the invasion of skin commensals and thus the occurrence of infection (figure 3a). An intact vasculature will help maintain the integrity and function of all these barriers and mechanisms. Deficiencies in skin integrity, immunity or vasculature can be considered risk factors for the development of cellulitis (figure 3b). Old age, diabetes and obesity cause defects in all three of these areas, and thus confer a relatively high risk. This combination of risk factors is often seen in patients with cellulitis who are hospitalised. The biggest risk factor, however, is a positive history for cellulitis.34
Old age comes with skin atrophy, poor circulation, immunosenescence, and comorbidities such as diabetes or congestive heart failure. Malnourishment causes impaired wound healing, decreased skin elasticity and integrity, and relative immunosuppression.35 Incidence, complication (e.g. bacteraemia, osteomyelitis, endocarditis) and hospitalisation rates are all higher in diabetic patients.53 Most cases of cellulitis in diabetic patients will be attributable to diabetic foot associated skin defects, but more than a quarter of the cases of culture-positive diabetic cellulitis occur on non-foot locations.54 In morbid obesity, the skin is more susceptible to damage and takes longer to repair.36
Some seasonal variability has been observed. Streptococcal skin infections occur more frequently in the winter in cold countries,55,56 while warmer regions see a higher erysipelas incidence during the summer.37
Skin microbiome alterations have been observed in diseases such as atopic dermatitis, where more staphylococci and fewer streptococci are present, but also in acne.57 S. aureus is shown to be overrepresented in the peri-abscess skin microbiome.58 Pioneering studies have revealed that commensals can influence the composition of the local microbiome and alter local immunity,59 but future studies will have to reveal relationships between the microbiome and cellulitis.
To admit, or not to admit
The 7% of patients who are hospitalised cause 83% of the total healthcare expenditure associated with cellulitis.6 Unfortunately, as yet there are no validated, prospectively evaluated admission guidelines. One system distinguishes classes with supposedly increasing mortality and therapy failure rates based on systemic symptoms, comorbidity and the Standardised Early Warning Scores.60,61 Two cohort studies compared this system with current clinical practice: one retrospectively, one prospectively.62,63
Overtreatment of infections that the system classified as mild (class I and II) was very common, while most of the severest infections (class IV) were undertreated. In one of the two studies, only 5 of 6 (83%) class IV patients had achieved complete resolution of symptoms at the end of therapy, compared with 100%, 98% and 96% in classes I-III.62,63 One explanation for this is that factors not incorporated in this system currently have a substantial effect on admission and treatment practices.
Pragmatically, one could consider admission for patients with (1) poor disease perception, (2) intake problems, (3) an altered mental status, or (4) disease progression despite adequately dosed oral antibiotics. Severity or dysregulation of comorbidity (e.g. diabetes, immunodeficiency, obesity, or cardiac, renal or venous insufficiencies) and severity of infection (e.g. systemic symptoms, organ failure) should also be taken into account.11,64
Factors predicting oral therapy failure may also be indications for admission for intravenous antibiotics. Retrospectively identified factors associated with failure of oral antibiotic therapy include fever, chronic leg ulcers, chronic oedema and lymphoedema, prior cellulitis in the same area, and wound infections.65,66 Additionally, after treatment in an observation unit for 24 hours, patients with cellulitis of the hand, an elevated lactate, fever, history thereof, or multiple comorbidities were more likely to be admitted.67,68 However, this mainly reflects clinical practice rather than need for admission.
Alternatively, outpatient parenteral antibiotic therapy, where intravenous antibiotics are given at home or on outpatient basis, can avoid or shorten hospitalisation for selected patients and is usually preferred by patients.69
One might wonder if a proportion of cases of cellulitis are self-limiting and do not require antimicrobial agents. It is noteworthy that in clinical trials performed in the pre-antibiotic era, in which the effects of horse serum and ultraviolet light were evaluated, cure rates of 70% were observed.70 On the other hand, it has also been demonstrated that inadequate empirical antibiotics are associated with prolonged treatment durations and length of hospital stay.71 Current treatment recommendations are summarised in figure 1.
Streptococci and S. aureus are the most common pathogens identified in patients with cellulitis (table 2), and accumulating evidence from prospective convalescent serology studies suggests that > 70% are caused by streptococci.4,79 Atypical pathogens can be observed in patients with selected conditions (table 3). In contrast to diabetic foot infections, diabetic non-foot infections are generally not caused by atypical pathogens.87 In the Netherlands, the preferred small spectrum agent covering both methicillin-susceptible S. aureus and beta-haemolytic streptococci is flucloxacillin. Confirmed streptococcal infections can be treated with benzylpenicillin or feneticillin. Co-amoxiclav and clindamycin are alternative options. Clindamycin is recommended in case of beta-lactam allergies, and inhibits streptococcal and staphylococcal toxin production. Clindamycin is also thought to have better tissue penetration than beta-lactams. However, clindamycin is highly concentrated intracellularly, and studies measuring tissue concentrations used homogenised tissues and thus also measured intracellular clindamycin.88 This overestimates relevant clindamycin levels in the extracellular fluid, while the primarily extracellular beta-lactam concentration is diluted by the released intracellular volume and thus underestimated.88 Of note, some S. aureus strains have inducible resistance for clindamycin, showing growth inhibition in vitro but resistance in vivo.89,90 In the Netherlands, around 10% of S. aureus from selected general practice patients and hospital patients show (inducible) resistance to clindamycin, compared with less than 3% for flucloxacillin.90 This makes clindamycin less preferable as an empirical choice.
Evidence does not favour one agent over others, although there is a major lack of evidence in this area.91 One study found pristinamycin to be slightly more efficacious than penicillin in a non-blinded trial, but did not account for penicillin not covering S. aureus. 3,92 Beta-lactams were as effective as non-beta-lactams in a cohort study.93 A recent meta-analysis comparing penicillins or cephalosporins with macrolides or lincosamides (such as clindamycin) found similar efficacy between the two groups.94
If one needs to cover multi-resistant Staphylococcus aureus (MRSA), vancomycin remains the first choice of treatment, with linezolid as an alternative.95 Additionally, three novel antibiotics have recently been approved by the European Medicines Agency for treatment of skin infections: oritavancin and dalbavancin, two (lipo)glycopeptides, and tedizolid, an oxazolidinone, all showing potent activity against MRSA similar to vancomycin and linezolid (table 4).96 Oritavancin and dalbavancin both have terminal half-lives of over two weeks and thus only require a single intravenous dose to reach cure rates non-inferior to a 2-week course of vancomycin.98,103 Whether this actually reduces the number of admissions or total treatment costs remains to be evaluated.
Optimising antibiotic use
For oral flucloxacillin, proper timing of intake (before or long after meals) optimises the bioavailability to ~55%.104 Beta-lactams reach lower serum concentrations in obese patients due to altered distribution volumes and clearance, so these patients might benefit from higher oral dosing, or more frequent intravenous dosing.105 This is underscored by the fact that obese patients tend to have lower cure rates.106,107
The optimal duration of antibiotic treatment of cellulitis is unknown. One study suggested that patients with cellulitis who are treated on an outpatient basis only require 5 days of therapy when signs of improvement are seen.108 However, this study used unconventional numbers for its power calculation, had a dropout rate of 30% before randomisation due to non-improvement, included relatively young and healthy subjects and made use of levofloxacin as study drug.11 Community-acquired pneumonia, pyelonephritis and intraabdominal infections require shorter antibiotic treatments than we previously thought necessary.109-111 Whether cellulitis treatment can also be shortened is under investigation.112
Some patients have an increased risk of a complicated infection. Obesity predisposes to local complications such as bullae, abscess formation, haemorrhagic lesions and necrosis.113,114 Smoking and delays in antibiotic treatment are also linked to abscess formation.113 Patients with congestive heart failure, neutropenia, hypoalbuminaemia, an altered mental status or discharge from the lesion have an increased risk of experiencing adverse outcomes, in terms of death, local complications (e.g. requiring surgical drainage) or systemic complications (e.g. multi-organ failure).7
Additional non-antibiotic management options can potentially improve outcomes. Compression therapy has long been, and still is, cause for debate. Advocates claim there is an accelerated reduction of oedema and pain, and shorter time to cure. Currently, no evidence supports this claim. Patients, however, often report side effects such as pain, dry skin, itching, constriction and slipping.115,116 It is unknown if the altered haemodynamics affect the time to microbiological cure. The adequacy of applied bandages varies in clinical practice, and inadequately applied bandages can cause pressure ulceration, thus unnecessary harm.117,118 An alternative to reduce oedema in the acute phase is passive leg elevation. To prevent persisting lymphoedema from causing recurrences, compression therapy is indicated when lymphoedema persists for several weeks after antibiotic treatment.11 Compression stockings should follow initial bandaging, provided the patient’s arterial disease status allows it.118
The use of anti-inflammatory drugs in addition to antibiotic therapy might be beneficial. In a proof-ofconcept study, adjunctive non-steroidal anti-inflammatory drugs (NSAIDs) led to faster regression and resolution of symptoms.119 Similarly, patients receiving adjunctive oral prednisone (2 days 30 mg, 2 days 15 mg, 2 days 10 mg, 2 days 5 mg) had earlier resolution of symptoms and intravenous to oral antibiotic switches.120,121 Whether these drugs also affect microbial eradication is as yet unknown.
Almost 30% of admissions for cellulitis are for recurrent cellulitis.122 Two, three and five year recurrence rates are 17%, 29-47% and 47%, respectively.38,123-125 Five-year recurrence rate is 57% in patients with a history of recurrence.125 For HIV-infected patients, one- and three-year recurrence rates are 29% and 47%.126 Independent of persisting risk factors that might explain recurrences, the first episode’s inflammation has also likely damaged local lymphatic channels. Drainage is then insufficient, antigen presenting cells cannot migrate, and accumulating protein-rich fluid accommodates invading bacteria.127
Lymphoedema is the most important risk factor for recurring cellulitis, and 25-60% of recurrent cellulitis patients suffer from chronic oedema.122,124 Obese patients have more recurrences and CRP and leucocyte counts are higher in these recurrences.39 For HIV-infected patients specifically, non-hepatitis liver disease, intravenous catheters or intravenous drug use increase the recurrence risk.126 All persisting risk factors are also likely to increase the chance of recurrences, and should be treated vigorously when possible. Lymphoedema warrants treatment with compression therapy. Tinea pedis should be treated with topical azoles in order to decrease the chance of recurrence. Frequent and meticulous interdigital web space cleansing prevents skin damage, bacterial overgrowth and bacterial invasion.127
S. pyogenes is able to survive and replicate within macrophages.128 Theoretically this might elicit recurrences. However, recurrence rates are similar between patients receiving antibiotics with or without intracellular activity.129 When infections recur despite adequately treating risk factors, prophylactic antibiotics prevent recurrences.130 In the PATCH I trial, which randomised 274 patients with two or more episodes to either twice daily low-dose oral penicillin or placebo, recurrence rates were significantly lower in the penicillin group (22% vs 37%) after one year, although this effect wore off after cessation of treatment.131 For recurring S. aureus infections, on-demand therapy can be considered. S. aureus eradication or hygiene measures do not prevent recurrent S. aureus skin infections.132,133
An overview of knowledge gaps which, if addressed, could advance our understanding of the pathophysiology of cellulitis and improve its clinical management is given in textbox 3. A major challenge is the high rate of misdiagnoses which can bias clinical trials towards non-inferiority.70 To determine applicability and reliability of trial results, it is imperative to document results from abscesses and cellulites separately, to accurately describe criteria and definitions, to extensively document clinical and microbiological characteristics, and to report information on additional procedures such as surgical drainage or limb immobilisation.134 For this relatively simple infection which has plagued humanity for so long, there still is a lot to discover.
We thank Dr. Harjeet Singh Virk for critically reviewing the manuscript.
The authors were supported by the Netherlands Organisation for Health Research and Development (ZonMW; grant no. 836011024). We have no conflicts of interest to report.