Introduction
Skin and soft tissue abscesses (SSTAs) comprise a number of pathologic entities with a spectrum of severity. Cutaneous abscesses involve the dermis and hypodermis and may be polymicrobial (containing bacteria found on nearby skin and mucosa) or due to Staphylococcus aureus (S. aureus) alone. Furuncles, the most common type of SSTA, develop spontaneously within hair follicles and are usually caused by S. aureus.Reference Stevens, Bisno and Chambers 1 Deeper soft tissue abscesses, including intramuscular abscesses, may be caused by injection drug use. The amount of cellulitis associated with SSTAs varies, but systemic toxicity is uncommon.
The late 1990s and early 2000s saw the worldwide emergence of community-associated methicillin-resistant S. aureus (CA-MRSA). While CA-MRSA can cause severe invasive infections, it predominantly causes SSTAs in young, otherwise healthy individuals.Reference Fridkin, Hageman and Morrison 2 , Reference Moran, Krishnadasan and Gorwitz 3 Between 1997 and 2005, outpatient visits in the United States for skin and soft tissue infections rose by 50% and more than doubled at emergency departments (EDs).Reference Hersh, Chambers and Maselli 4 In 2008, CA-MRSA accounted for 59% of all culturable skin and soft tissue infections in the U.S. EDs, 85% of which were abscesses.Reference Talan, Krishnadasan and Gorwitz 5
Current guidelines from the Infectious Disease Society of America (IDSA) emphasize that the primary treatment for SSTAs, including those due to CA-MRSA, is incision and drainage.Reference Liu, Bayer and Cosgrove 6 According to such guidelines, adjunctive antibiotics should be reserved for complicated abscesses. Among features indicating complicated infection, the presence of associated cellulitis, in particular, provides a reason to prescribe antibiotics in many cases, because some induration and erythema surround most SSTAs.
Evidence on the potential benefit of adjunctive antibiotics, in addition to incision and drainage, in the treatment of uncomplicated SSTAs consists of several small, randomized controlled trials (RCTs) and data from observational studies.Reference Fridkin, Hageman and Morrison 2 , Reference Moran, Krishnadasan and Gorwitz 3 , Reference Macfie and Harvey 7 – Reference Davis, Perri and Donabedian 8 Although most RCTs demonstrate no benefit, a few show a trend toward benefit,Reference Macfie and Harvey 7 , Reference Singer and Thode 11 and some authors have argued that these studies were underpowered to show a benefit that likely exists.Reference Spellberg, Boucher and Bradley 9 A 2007 systematic review and 2013 meta-analysis found no benefit from adjunctive antibiotics.Reference Hankin and Everett 10 , Reference Singer and Thode 11 Both studies had important limitations. The 2007 systematic review did not include data from two recent RCTs and four observational studies. Many trials included in the review were conducted before the emergence of CA-MRSA, and it is possible that SSTAs due to CA-MRSA are unique and benefit more from antibiotics. The meta-analysis was limited to only RCTs and missed one important RCT and all observational study data. The report also failed to address the possibility that SSTAs caused by CA-MRSA represent a distinct entity. Although many of the observational studies of SSTA treatment have been conducted since the emergence of CA-MRSA, data from these studies have yet to be appraised in combination with RCTs.
Lacking robust, contemporary evidence to guide clinical practice, and possibly concerns over the potential for CA-MRSA to cause more severe infections, it seems that clinicians continue to err on the side of prescribing antibiotics. Presented with a hypothetical case of an uncomplicated SSTA, a non-random sample of polled clinicians worldwide would prescribe adjunctive antibiotics.Reference Hammond and Baden 12 In the United States, emergency physicians prescribe antibiotics for roughly 75% of abscess/cellulitis cases, increasingly choosing an antibiotic active against MRSA.Reference Pallin, Egan and Pelletier 13 As the prevalence of these infections has increased, so has the number of antibiotic prescriptions.
Given continued uncertainty about the role of antibiotics in the treatment of SSTAs, we saw a need for a systematic appraisal of the relevant literature, including observational studies. Our objective was to assess the effect of adjunctive antibiotics, in addition to incision and drainage on treatment success in uncomplicated SSTAs.
Methods
Study design
We performed a systematic review and meta-analysis of the effect of antibiotics on the treatment of uncomplicated SSTAs. Our study conforms to PRISMA Statement guidelines for systematic reviewsReference Moher, Liberati and Tetzlaff 14 and MOOSE guidelines for observational studies.Reference Stroup, Berlin and Morton 15 Our study protocol is available upon request.
Search strategy and selection criteria
In conjunction with a medical librarian, we performed a comprehensive literature search of the following databases: MEDLINE (1950 to December 2013), EMBASE (1974 to December 2013), Web of Knowledge (1970 to December 2013), and Google Scholar (no date restriction). Our search strategy and terms can be found in Appendix 1. All languages were included. We included only studies of primary data.
We searched ClinicalTrials.gov to identify unpublished trials and contacted investigators in an attempt to obtain any pilot, preliminary, or otherwise unpublished data. We reviewed reference lists of studies, review articles, and editorials, and consulted with topic experts, to identify additional studies not retrieved by the search.
We included RCTs and observational cohort studies from ED or ambulatory care settings, with adult and pediatric subjects. We included RCTs that satisfied the following criteria: 1) patients with SSTAs who underwent incision and drainage, 2) treatment group that included any antibiotics, 3) control group that included placebo, and 4) treatment failure or success that was ascertained on follow-up. For observational studies, the criteria were 1) patients with SSTAs who underwent incision and drainage, 2) treatment group that included appropriate antibiotics (defined in the section on interventions and outcomes), 3) control group that included no antibiotics or inappropriate antibiotics (defined in the section on interventions and outcomes), and 4) treatment failure or success that was ascertained on follow-up. We excluded studies where the abscess cavity was sutured closed. Our rationale was that, because most studies retrieved by our search involved incision and drainage alone (as recommended by the IDSAReference Liu, Bayer and Cosgrove 6 ), by including studies of primary suture closure, we would introduce another significant source of heterogeneity while adding relatively few subjects to the analysis. We also excluded studies of postsurgical wounds.
Screening and data abstraction
Two authors (Fahimi and Singh) reviewed abstracts retrieved by the initial search strategy with respect to the inclusion and exclusion criteria. Full articles were reviewed when one or both authors deemed the abstract appropriate. The authors independently assessed suitability of studies for inclusion in the meta-analysis and reached complete agreement. When published data were incomplete or did not provide sufficient information for pooling, we attempted to contact the authors for additional data that would allow inclusion. Additional data from author correspondence were included after all three authors agreed. For each study, sample sizes, treatment allocation, and treatment success were extracted by two authors (Fahimi and Singh) and determined by consensus.
Interventions and outcomes
For RCTs, we defined our intervention group as patients who received an antibiotic and our control group as patients who received either no antibiotic or placebo. For observational studies of patients with culture-positive abscesses, appropriate therapy was defined as an antibiotic to which the bacterial isolate was susceptible in vitro, and inappropriate as one to which the isolate was resistant. We adhered to an intention-to-treat approach and report medication adherence for each arm when available from the data. Our primary outcome was treatment success, that is, lack of treatment failure. Treatment failure was defined as persistence of the primary infection at the end of the treatment period or need for an additional intervention or procedure. Although this outcome definition is broad, our inability to differentiate types of treatment failure stems from nonuniform outcome definitions and limited data presented in the primary studies.
For this analysis, we assumed that the following variables would not significantly affect pooled results: study setting, identification and selection of patients with an abscess requiring drainage, drainage technique, duration of antibiotic administration, and post-procedural care.
Quality scoring
Two authors (Fahimi and Singh) assessed the quality of included studies. RCTs were appraised according to guidelines published by the Cochrane Collaboration.Reference Higgins and Green 16 For observational studies, we applied the Newcastle-Ottawa Scale quality assessment tool.Reference Wells, Shea and O’Connell 17 There was complete agreement on quality assessment.
Meta-analysis
We calculated relative risks (RR) of treatment success with 95% confidence intervals (CI) for each study. Our primary analysis pooled all RCTs comparing antibiotics to placebo. We performed two secondary analyses: 1) effect of appropriate antibiotics (defined as those to which the isolate proved susceptible in vitro) compared to inappropriate antibiotics (to which the isolate was not susceptible in vitro) or no antibiotics, in observational studies of patients with confirmed MRSA infection (termed culture-proven MRSA-only analysis); and 2) effect of antibiotics compared to placebo in RCTs, or appropriate antibiotics compared to inappropriate or no antibiotics in observational studies, restricted to studies in which all patients were enrolled after 1995 (termed MRSA-era analysis).Reference Herold, Immergluck and Maranan 18 This analysis assumed nonuniform prevalence of MRSA across study populations or geographic areas. A summary of the analyses, including study types and comparison groups, can be seen in Appendix 2.
The MRSA-era analysis was an attempt to include a population of subjects and SSTAs with MRSA prevalence similar to that encountered today. Combining data from both RCTs and cohort studies to meta-analyse the effect of an intervention is well supported and may provide significant advantages over pooling results from RCTs only.Reference Shrier, Boivin and Steele 19 , Reference Golder, Loke and Bland 20 Our intervention categories, appropriate and inappropriate antibiotics, based on in vitro susceptibility data in observational trials, correspond with, but admittedly are not equivalent to, antibiotic therapy and placebo in RCTs. Grouping together interventions that were not precisely equivalent was necessary, however, in order to allow pooling of data between observational studies and between observational studies and RCTs.
We assessed for heterogeneity by the χ2 test statistic. From this, we calculated the I 2 statistic (I 2=100% x [χ2–df]/χ2), which is the percentage of total variation across studies due to heterogeneity rather than chance. Traditionally, cut-offs of 25%, 50%, and 75% correspond respectively to low, moderate, and high levels of heterogeneity,Reference Higgins, Thompson and Deeks 21 though some authors consider high heterogeneity to be at I 2 levels greater than 50%.Reference Hatala, Keitz and Wyer 22 Negative I 2 values were put equal to zero, suggesting no observed heterogeneity. If no heterogeneity was present, we used a fixed effects model as described by GreenlandReference Greenland 23 to calculate the pooled RR. When studies were found to be heterogeneous (χ2>df), we report results from a random effects model and, secondarily, from a fixed effects model with the 95% CI recalculated using the adjustment described by Shore et al., where between-study heterogeneity is incorporated into calculations of variance.Reference Shore, Gardner and Pannett 24 , Reference DerSimonian and Laird 25 Publication bias was evaluated through visual inspection of funnel plots, as well as through both Egger’sReference Egger, Davey Smith and Schneider 26 and Begg’sReference Begg and Mazumdar 27 tests for small-study effects. We report the kappa statistic for inter-reviewer agreement for inclusion of studies. Statistical analysis was done using Microsoft Excel (Microsoft, Redmond, WA) and Stata 11.1 (StataCorp, College Station, TX).
Results
We screened 1,968 records retrieved in the computerized search and 10 studies identified by reviewing reference sections and through discussion with topic experts. We contacted the investigators of four ongoing trials but were unable to obtain any preliminary, pilot, or otherwise unpublished data. Inter-reviewer agreement for inclusion of records (title and abstract) was excellent (kappa=0.88).
We reviewed the full text of 19 studies (7 RCTs and 12 observational studies). Seven studies were excluded after review. There was complete agreement between reviewers on which of these studies fulfilled inclusion criteria (kappa=1.0). Details of the excluded studies are highlighted in Appendix 3. In three observational studies, the data could not be directly pooled. We contacted the authors who provided additional information allowing us to include the studies in the meta-analysis.Reference Moran, Krishnadasan and Gorwitz 3 , Reference Higgins and Green 16 , Reference Wells, Shea and O’Connell 17 We ultimately included 12 studies (5 trials and 7 observational studies) in our analyses with a total of 1,969 subjects. Our study selection and inclusion process are outlined in Figure 1.
Figure 1 Outline of study selection and inclusion.
Characteristics of studies
The primary meta-analysis included five RCTs. One trial, done in 1977, lacked clear description of the methodologyReference Macfie and Harvey 7 (Table 1). The remaining four trials used proper randomization and blinding techniques and were free of selective reporting bias. The majority suffered from incomplete outcome data due to high lost to follow-up rates.Reference Davis, Perri and Donabedian 8 – Reference Singer and Thode 11 Additional sources of bias were inclusion of patients on antibiotics prior to enrollment,Reference Macfie and Harvey 7 and use of beta-lactam antibiotics despite a high rate of CA-MRSA in the study population.Reference Spellberg, Boucher and Bradley 9 Four of the RCTs were placebo-controlled.Reference Davis, Perri and Donabedian 8 – Reference Singer and Thode 11 Treatment failure was defined as the need for repeat incision and drainage and/or need for antibiotics, except in one trial where it was defined as persistent infection.Reference Davis, Perri and Donabedian 8 Adherence with therapy was not assessed in three trials.Reference Macfie and Harvey 7 , Reference Spellberg, Boucher and Bradley 9 , Reference Singer and Thode 11 In the remaining two trials, adherence with therapy was generally poor.Reference Davis, Perri and Donabedian 8 , Reference Hankin and Everett 10 We noted a wide range in reported rates of treatment failure between trials (median 4.1%, range 0%−17% in the antibiotic group and median 5.3%, range 3.7%−26.5% in the no antibiotic group). Details of the included RCTs are presented in Table 2.
Table 1 Assessment of bias in randomized controlled trials
+=Low risk of bias; −=high risk of bias;? =uncertain risk of bias.
Table 2 Randomized controlled trial methodology and highlights
BID=twice a day; ED=emergency department; I&D=incision and drainage; PO=per os (by mouth); SSTA=skin and soft tissue abscess; TMP-SMX=trimethoprim/sulfamethoxazole.
The seven observational studies included five cohort studies—one prospectiveReference Hammond and Baden 12 and four retrospectiveReference Pallin, Egan and Pelletier 13 – Reference Stroup, Berlin and Morton 15 , Reference Wells, Shea and O’Connell 17 —that primarily assessed antibiotic effect on treatment success. Data from subgroups of two studies that did not primarily examine antibiotic use were also included. One of these studies was a trial of abscess drainage techniques,Reference Higgins and Green 16 and the other study was a large cohort study of skin and soft tissue infections.Reference Moran, Krishnadasan and Gorwitz 3 The use of appropriate antibiotics was compared to inappropriate antibiotics in four studies,Reference Pallin, Egan and Pelletier 13 – Reference Stroup, Berlin and Morton 15 , Reference Wells, Shea and O’Connell 17 to no antibiotics in two studies,Reference Hammond and Baden 12 , Reference Higgins and Green 16 and to inappropriate or no antibiotics in one.Reference Moran, Krishnadasan and Gorwitz 3 Details of these studies are shown in Table 3, and a quality assessment summary is shown in Table 4.
Table 3 Observational studies, methodology, and highlights
* Appropriate antibiotics is defined as those to which isolate was susceptible in vitro; inappropriate antibiotics is defined as those to which isolate was not susceptible in vitro.
ED=emergency department; I&D=incision and drainage; NA=not applicable; SSTA=skin and soft tissue abscess.
Table 4 Assessment of bias in observational studies
+=Low risk of bias; −=high risk of bias;? =uncertain risk of bias.
Of the 12 included studies, only 6 reported loss to follow-up.Reference Davis, Perri and Donabedian 8 – Reference Singer and Thode 11 , Reference Pallin, Egan and Pelletier 13 , Reference Moher, Liberati and Tetzlaff 14 Among those who did report, not all classified from which arm that the loss came. Otherwise, no evidence for differential loss to follow-up was found. All but one of the studies reported no financial disclosures. In that case, multiple authors were on advisory boards of pharmaceutical companies.Reference Moran, Krishnadasan and Gorwitz 3
Results of primary analysis
None of the five RCTs in the primary meta-analysis showed a significant benefit to antibiotic treatment. The pooled RR of treatment success with antibiotics using a random effects model was 1.03 (95% CI 0.97−1.08) (Figure 2). Tests of heterogeneity resulted in χ2=5.81 (p<0.21), with an I 2 =31.2%, corresponding to moderate heterogeneity. Differences among studies that contribute to heterogeneity include the year that they were conducted (before or after the emergence of CA-MRSA), study setting (e.g., pediatric ED v. surgical clinic), types of antibiotics used, exact definition of treatment failure, method of follow-up (direct wound inspection v. phone follow-up), and assessment of antibiotic adherence. A funnel plot, as well as both the Egger’s and Begg’s tests, showed no evidence of publication bias.
Figure 2 Randomized controlled trials forest plot. Note: RR>1 favors the use of antibiotics.
Secondary analyses
In the four studies limited to culture-proven MRSA-only infections, the pooled RR of treatment success with appropriate antibiotics (defined previously), as compared to inappropriate or no antibiotics, was 1.05 (95% CI 0.96−1.15). When the results of the three randomized trials and three observational studies conducted in the MRSA-era were combined (using the fixed effects model because there was no observable heterogeneity), the pooled RR of treatment success with antibiotics/appropriate antibiotics was 0.99 (95% CI 0.98−1.01). This result was unchanged after excluding the single RCT that used an antibiotic known to be inactive against MRSAReference Spellberg, Boucher and Bradley 9 (data are not shown). The prevalence of MRSA in the MRSA-era studies ranged from 34% to 80%. A summary of these results, including measures of heterogeneity, is presented in Table 5 and Figure 3. Funnel plots for all analyses and all 12 included studies can be found in Appendix 4. One studyReference Hammond and Baden 12 was not ultimately included in any of the three quantitative meta-analyses because it was an observational study and predated the emergence of MRSA. In this study, there were no treatment failures in any either cohort (antibiotics or no antibiotics).
Figure 3 Summary of pooled analyses. This figure shows the pooled RR and 95% CI for the primary analysis and each secondary analysis. Note: RR>1 favors the use of antibiotics; RR<1 favors placebo, no antibiotics, or inappropriate antibiotics—described here as “favors no antibiotics.” *Fixed effects model results.
Table 5 Summary of studies included in each analysis with pooled results
* Pooled analyses with negative I 2 values were set to zero, suggesting no observed heterogeneity; only fixed effects model results were reported.
† MRSA-only subgroup of this study was used in this analysis.
CI=confidence interval; NA=not applicable.
Discussion
Our systematic review found no evidence to support the routine use of antibiotics in addition to incision and drainage for the treatment of uncomplicated skin and soft tissue abscesses. The RR point estimates for treatment success in our three analyses range from 0.99 to 1.05, although heterogeneity among studies and wide CIs in the RCT-only and MRSA-only analyses may limit the strength of our conclusions. Our results reinforce the findings of two previous, smaller reviews, while focusing particularly on CA-MRSA infections.Reference DerSimonian and Laird 25 , Reference Egger, Davey Smith and Schneider 26 Our findings support current IDSA guidelines, which recommend incision and drainage alone for uncomplicated cutaneous abscesses, while acknowledging the need for more data.Reference Liu, Bayer and Cosgrove 6
Our two secondary analyses attempt to address the question of whether antibiotics may be of particular benefit when CA-MRSA is a likely cause of the infection. This is an important question because CA-MRSA now causes a large proportion of community-associated SSTAs in the United StatesReference Talan, Krishnadasan and Gorwitz 5 and a substantial, though more variable, proportion in Europe and Asia.Reference Chua, Laurent and Coombs 28 Regardless of the exact clonal type, CA-MRSA is genotypically and likely phenotypically distinct from methicillin-sensitive S. aureus. The USA 300 strain, for example, although predominantly associated with furuncles and cutaneous abscesses, is also increasingly the cause of invasive disease in the United States.Reference Fridkin, Hageman and Morrison 2 CA-MRSA may represent a reason to treat even simple abscesses with antibiotics active against MRSA, or at least to lower the threshold for antibiotic treatment. In this light, it is significant that neither our MRSA-only or MRSA-era analyses found a statistically significant benefit from antibiotics. However, the MRSA-only analysis did find an RR point estimate of 1.05, with an upper confidence limit of 1.15, suggesting that the number needed to treat to prevent one treatment failure could be as low as seven and underscoring the need for more data.
Our study assesses the effect of adjunctive antibiotics on one important outcome; however, there are other potentially important outcomes not addressed, including abscess recurrence and antibiotic-related harms. With respect to recurrence, two trials found a significant reduction in new cutaneous abscesses (predominantly nearby “satellite” lesions) in the antibiotic group at 7 to 30 days, although this was not the primary outcome.Reference Hankin and Everett 10 , Reference Singer and Thode 11 Potential harms of antibiotics include rashes and severe allergic reactions; gastrointestinal side effects, including Clostridium difficile-associated diarrhea; yeast infections; and drug-drug interactions. Antibiotics cause 19% of ED visits for drug-related complications, mostly allergic reactions, and sulfonamides and clindamycin are associated with the highest risk.Reference Shehab, Patel and Srinivasan 29 One trial included in our analysis reported adverse events in 20% of subjects treated with trimethoprim-sulfamethoxazole.Reference Hankin and Everett 10 More broadly, the development of resistance from increasing or inappropriate use of antibiotics has been identified as a major threat to public health and a reason for research to identify unnecessary antimicrobial therapies.Reference Schwartz, Bell and Hughes 30 , Reference Wise, Hart and Cars 31
Our study has a number of limitations inherent in meta-analysis, which we attempted to address. The most important of these is that the strength and validity of the overall conclusions depend on the number, size, and quality of included studies, and is limited by heterogeneity among studies. In order to capture all potentially relevant studies, we used an extensive search strategy, included multiple databases, and contacted authors directly to obtain additional data. The small number of included trials and total number of subjects may be insufficient to demonstrate a small statistically significant benefit (or harm) from antibiotics, which actually exists. We tried to strike an appropriate balance between including as many studies as possible and excluding those with quality issues. Some quality issues persisted in individual studies, including the enrolment of a nonconsecutive convenience sample (i.e., potential selection bias), attrition (i.e., incomplete outcome data bias), subjectivity of what constitutes a treatment failure (i.e., outcome assessment bias), and the use of telephone follow-up.
The problem of subjects lost to follow-up pervades many of the studies. The proportion lost to follow-up varies from 2% to 38%. If all subjects lost to follow-up in the antibiotic arms were cured and all those in the nonantibiotic arms had a treatment failure, the analysis might have found a statistically significant benefit for antibiotics. It is more likely, however, that most patients lost to follow-up in both arms were cured (because patients with persistent symptoms or complications are more likely to return), thus minimizing the potential effect of patients lost to follow-up on overall results.
Heterogeneity among included studies was a significant problem. Although we attempted to account for measured heterogeneity by using both a random effects model and a fixed effects model with statistical correction in calculations of variance, unmeasured (clinical) heterogeneity was difficult to overcome. It is important to understand that, in SSTA studies, determinations regarding enrolment and exclusion criteria, as well as outcomes, are unavoidably subjective. These determinations likely varied by study setting, depending on whether the study population was pediatric or adult and whether providers were surgeons, internists, or emergency physicians. The type and severity of infections and the reported outcomes likely varied systematically among studies. For example, infections included in a study from a pediatric EDReference Hankin and Everett 10 (and those excluded because they required admission) were likely different from those from an urban, public hospital surgical clinic serving injection drug users.Reference Spellberg, Boucher and Bradley 9 Observational studies comparing inappropriate to appropriate antibiotics tended to enrol abscesses severe enough to be treated with antibiotics in normal practice,Reference Moran, Krishnadasan and Gorwitz 3 , Reference Moher, Liberati and Tetzlaff 14 whereas other studies seemed to select relatively minor abscesses.Reference Singer and Thode 11 , Reference Hammond and Baden 12 Differences in antibiotic regimens used are another source of heterogeneity. Additional potential sources include differences in bacterial culture methods and sensitivity criteria used that were not uniformly reported. Although these many sources of clinical heterogeneity limit our pooled results and weaken our conclusion, we believe that our logically constructed secondary analyses, with careful attention to meta-analytic methods and detailed description of individual studies, should allow readers to draw independent conclusions.
The heterogeneity we observed among SSTA studies also reflects the fact that the disease itself is extremely heterogeneous. The term uncomplicated SSTA comprises a number of pathologic entities, each with a spectrum of severity. Examples include simple furuncles, cutaneous abscesses that extend down into the pannus of the buttock or thigh, cutaneous abscesses that have a substantial patch of surrounding cellulitis, and intramuscular abscesses due to injection drug use. Such a spectrum presents difficulties for both researchers and clinicians. For researchers, it makes it difficult to define and standardize inclusion criteria and determine treatment success or failure (i.e., whether further drainage is needed or whether persistent cellulitis is significant) among a range of infection types.
For practicing clinicians, it is difficult to apply uniform management to a fundamentally heterogeneous group of infections, particularly in the absence of evidence-based guidelines on how to gauge SSTA severity or identify an abscess as “complicated.” For this reason, it is understandable that treatment decisions (i.e., no antibiotics, oral antibiotics, or parenteral antibiotics) and disposition (i.e., discharge, admission, or surgical consultation) have tended to err on the conservative side. In this light, our analysis of treatment trials and observational studies—which spans a range of study settings, infection types (including MRSA), and severity—showing that drainage alone, without adjunctive antibiotics, is sufficient treatment for uncomplicated SSTAs, should be reassuring to clinicians. Our results should further empower clinicians to say “no” to adjunctive antibiotics, with their associated expense and potential for harm.
Our results support a recommendation that treatment of uncomplicated SSTAs, even when MRSA is a likely etiology, should be with incision and drainage alone. Antibiotics should be reserved for special circumstances indicating a complicated infection, in accordance with IDSA guidelines. These circumstances include extensive disease, rapidly spreading associated cellulitis, associated phlebitis or lymphangitis, fever, co-morbidities and immunosuppression, extremes of age, and lack of response to drainage alone.Reference Liu, Bayer and Cosgrove 6 We recognize that, even in the absence of such circumstances, the term uncomplicated leaves clinicians with some latitude, particularly regarding surrounding cellulitis. Clinicians may choose to practice conservatively while awaiting more robust data.
The future research we look forward to seeing in this area includes large, multicentre trials comparing placebo to antibiotics active against CA-MRSA, with clearly defined enrolment and outcome criteria. Such trials are ongoing. 32 , 33 Prospective observational studies and subgroup analyses from large trials are also needed to identify clinical features associated with adverse infection-related outcomes. Ultimately, physicians on the front-lines will need simple clinical decision rules to distinguish the small subgroup of patients with SSTAs who may benefit from adjunctive antibiotics from the vast majority who do not.
Competing interests: None declared.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/cem.2014.52
