Comparative efficacy of antibiotics for the treatment of acute bacterial skin and skin structure infections (ABSSSI): a systematic review and network meta-analysis

Abstract

Objective:

The objective was to conduct a systematic review and network meta-analysis (NMA) of existing treatments for ABSSSI focusing on the novel lipoglycopeptide oritavancin.

Methods:

EMBASE, MEDLINE, MEDLINE in Process, CENTRAL (Cochrane), and select conferences were searched for randomized controlled trials investigating antimicrobial agents for the treatment of ABSSSI. NMA was used to estimate the odds ratios of the Test-Of-Cure (TOC) and Early Clinical Response (ECR) outcomes for treatments relative to vancomycin in the ITT populations. Sub-group analyses in MRSA and MSSA populations were conducted for TOC; sensitivity analyses investigated the use of the clinically evaluable (CE) populations and the restriction to trials following the recent FDA guidelines for clinical trials.

Results:

The systematic review identified 52 trials. The most commonly investigated treatments were vancomycin and linezolid; most trials reported TOC, but not ECR. The posterior mean and 95% credible intervals for odds ratios of TOC for antimicrobial agents relative to vancomycin were: linezolid (1.55; 0.91–2.57), daptomycin (2.18; 0.90–5.42), and oritavancin 1200 mg (1.06; 0.80–1.43). The odds ratio of ECR for oritavancin 1200 mg was 1.02 (0.23–4.33). In the MRSA sub-group the odds ratios relative to vancomycin for TOC were: linezolid (1.55; 0.96–2.46), daptomycin (0.74; 0.13–3.66), and oritavancin 1200 mg (0.94; 0.44–2.02). In the MSSA sub-group they were linezolid (1.36; 0.15–13.34) and oritavancin 1200 mg (0.82; 0.08–7.83). These results were robust to the sensitivity analyses.

Conclusions:

This NMA provides a unified framework for the comparison of all available antimicrobial agents used in the treatment of ABSSSI and is the first to assess the ECR end-point. The results suggest equivalence of clinical efficacy between vancomycin, daptomycin, linezolid, and novel antimicrobial agents including oritavancin for the treatment of ABSSSI at TOC. The wide uncertainty margins indicate the heterogeneity of the available evidence and the need for further research.

Keywords: :

• ABSSSI
• Oritavancin
• Systematic review
• Network meta-analysis
• Test-of-cure
• Early clinical response

Introduction

Acute bacterial skin and skin structure infections (ABSSSI) incidence is rising globally1, with recent annual incidence estimates requiring healthcare intervention in the US of 4.96%2, which represents an increased healthcare and economic burden3^,4. The 2010 FDA guidelines define ABSSSI as bacterial infections of the skin with a lesion size of at least 75 cm² (measured by the area of redness, edema, or induration) and include cellulitis, wound infection, and abscesses with surrounding cellulitis5. ABSSSI are caused by many different bacteria, most commonly gram-positive bacteria including methicillin-resistant Staphylococcus aureus (MRSA) and methicillin-susceptible S. aureus (MSSA)6^,7 and with increasing antibiotic resistance treatment is becoming more difficult8–10. A surveillance study found 44.6% of ABSSSI in North America were caused by S. aureus and 80% of these exhibited methicillin resistance11. Current treatment can include administration of broad-spectrum antibiotics; however, in the case of known or suspected MRSA targeted antibiotics are administered over 7–14 days12–14.

Vancomycin, linezolid, and daptomycin are among the most common antibiotics used in the treatment of ABSSSI, but there is limited evidence on their comparative effectiveness15^,16. Recently approved antibiotics for the treatment of ABSSSI, such as oritavancin17, dalbavancin18, and tedizolid19, further highlight the need for comparative effectiveness research. The majority of published randomized controlled trials (RCTs) have been compared to vancomycin and have not achieved statistically significant differences15, although, arguably, the strongest results have been found in favor of linezolid over vancomycin20–23. Thirteen meta-analyses have been conducted comparing the efficacy of vancomycin, linezolid, and other anti-microbial agents, the majority comparing linezolid with vancomycin and finding clinical superiority of linezolid7^,12^,24–34. However, most of these considered only direct comparisons of two treatments and neglected the impact of indirect evidence.

Bayesian network meta-analysis (NMA) is a statistical method used to combine the results from multiple studies and indirectly compare multiple treatments, thus incorporating all available direct and indirect evidence35–37. Two Bayesian NMAs have been conducted in ABSSSI caused by MRSA. The first compared telavancin, daptomycin, vancomycin, linezolid, tigecycline, and dalbavancin and found higher success rates for linezolid, dalbavancin, and telavancin, but the analysis only included studies published by 2010, and omitted recently approved antimicrobial agents such as oritavancin7. The second was conducted in 2012 and compared vancomycin, linezolid, and ceftaroline26. Since the publication of these analyses, the FDA has issued guidance on the conduct of clinical trials in ABSSSI, and studies following these guidelines are expected to be more homogeneous and appropriate for NMA5.

These limitations of the current evidence on the comparative efficacy of antimicrobial agents, particularly with regards oritavancin and other novel therapies for the treatment of ABSSSI, motivates the conduct of a systematic literature review and Bayesian NMA.

Methods

A systematic literature review and Bayesian NMA was conducted to compare the efficacy and safety of common antimicrobial agents used in the treatment of ABSSSI.

Systematic literature review

RCTs meeting inclusion criteria were identified by searching the EMBASE, MEDLINE, and MEDLINE in Process and the Cochrane Central Register of Controlled Trials databases. Searches were conducted from database inception to January 2, 2014 (available on request). In addition, conference proceedings from the 2012 and 2013 Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), Infectious Diseases Week (ID Week), and European Society for Clinical Microbiology and Infectious Diseases (ECCMID) conferences were searched for relevant abstracts describing RCTs meeting inclusion criteria.

Publications were included if they recruited adult patients with ABSSSI and if the investigated interventions included oritavancin, vancomycin, linezolid, tigecycline, dalbavancin, daptomycin, ceftaroline, teicoplanin, clindamycin, or telavancin; additional treatments were included if they were investigated in a study that included one of the listed interventions. The following outcomes were considered: clinical response, clinical cure, Early Clinical Response (ECR), lesion size reduction, relapse/recurrence, sustained clinical response, overall incidence of adverse events (AEs), treatment emergent AEs, serious AEs, and the most commonly reported AEs across studies (nausea, vomiting, diarrhea, pain, fever, headache, and pruritus). Studies were restricted to English language RCTs.

The risk of bias of the included full publications was assessed using the Cochrane risk of bias tool38. Important aspects of risk of bias are not normally reported in conference abstracts, so a critical appraisal of trials reported in abstracts only was not conducted.

Definitions and outcomes

The FDA guidance for ABSSSI recommends four study populations for analyses. The primary population that should be considered is the intention to treat (ITT) population, which includes all randomized patients; other populations are the clinically evaluable (CE) population, which includes patients from the ITT population who follow important components of the trial such as attendance at follow-up examinations and demonstrate compliance to treatment (definitions vary by study), the microbiologically evaluable (ME) population which includes ITT patients with a confirmed baseline bacterial pathogen, and the safety population which includes all patients who received at least one dose of study drug.

Base case analyses were conducted using trials reporting efficacy of antimicrobial agents in the treatment of ABSSSI due to a mixture of microbial causes. Two key outcomes were chosen for these analyses. The first is treatment success at the ‘Test-Of-Cure’ (TOC) visit, commonly 7–14 days after the end of treatment, in the intention to treat (ITT) population (as recommended by the FDA guidance). Treatment success in ABSSSI can be determined either clinically or microbiologically. Clinical success is generally characterized as resolution of the symptoms associated with the infection present at admission or improvement to the extent that no further therapy was necessary; in the absence of a culture it is presumed that microbiological eradication has occurred. When a culture is taken, microbiological success is defined as eradication or presumed eradication of the baseline pathogen in a culture taken at a follow-up visit. These measures of success were judged equivalent for the purposes of this analysis. The second outcome was ECR. As of October 2013, this is the primary end-point for the clinical assessment of treatment success in ABSSSI of the FDA in the US and is defined as cessation of spread of skin lesion erythema and absence of fever at 48–72 h after treatment initiation5.

Sub-group analyses were performed for the microbiologically confirmed MRSA and MSSA populations for the TOC and ECR outcomes. By definition, these were conducted in the microbiologically evaluable (ME) populations.

Sensitivity analyses assessing robustness of patient population definitions were conducted. To clinicians, the response of patients who are compliant to treatment (i.e. CE population) can be seen as more informative than response in the ITT population. Therefore, analyses of the TOC and ECR end-points were conducted using the CE population in preference to the ITT population. In addition, the TOC end-point in the ITT and CE populations using the sub-set of trials that followed recent FDA guidance on patient inclusion in clinical trials were analyzed5. This guidance recommends including patients with the previously noted FDA definition of ABSSSI and a mixture of both male and female patients and ABSSSI types, such as wound infection or major cutaneous abscess. A number of patient types to exclude are also recommended, such as those who have received more than 24 h of successful treatment for the current episode of ABSSSI or those with other medical conditions, for example neutropenia, that may interfere with the interpretation of the primary end-point. This reduces the heterogeneity of ABSSSI patients across trials.

Statistical analysis

The Bayesian NMA was conducted using recommendations from the NICE Decision Support Unit for count data modeling, and adopted a generalized linear model with a binomial outcome and logistic link function39. Vague priors were used for baseline and treatment effects. Treatment effects were estimated as mean odds ratios compared to vancomycin, alongside Bayesian credible intervals (CrI). The model code is provided in the Supplementary Appendix.

Random treatment effects (RE) models were used to account for heterogeneity across trials40. The Deviance Information Criterion (DIC)41 was used to compare the model fit. A random effects model was used for each analysis unless its DIC was five units higher than that of the fixed effects model42. The exception is if less than three studies were included in the analysis or if the posterior estimates were unstable, in which case the fixed effects model was preferred.

Inconsistency between direct and indirect treatment effect estimates may also be present in an evidence network43^,44. The consistency between direct and indirect evidence was assessed for the best fitting model for each end-point using the node-splitting approach45.

Analyses were conducted using OpenBUGS (Version 3.2.2 rev 1063 2012-07-15, Medical Research Council and Imperial College, London, UK)42^,46. Samples from 50,000 iterations of the posterior distribution of two chains, with a 50,000 iteration burn-in to ensure convergence, were used for parameter estimates. Caterpillar and Brooks-Gelman-Rubin (BGR) plots of the estimated parameters were examined to ensure that the models converged satisfactorily42.

Results

Results of systematic literature review

The systematic review identified 52 trials in 50 published articles (46 trials) and eight conference abstracts (six trials) (Figure 1). Two of the six trials identified in conference abstracts were found to have been published by June 5, 2014 and the clinical study reports for SOLO I3 and SOLO II17 were provided by The Medicines Company. Vancomycin and linezolid were the most frequently reported treatments, being included in 26 and 18 out of the 52 trials, respectively.

Figure 1. PRISMA flowchart illustrating systematic literature review strategy and article selection process.

The trials were generally judged to be at low risk of bias (results presented in the Supplementary Appendix). An exception was in studies where TOC was assessed clinically but the assessors were inadequately blinded. Treatment doses and administration across trials were considered clinically homogenous. Two studies were excluded from the analysis for administering doses not used in clinical practice; the daptomycin arm of Katz et al.47 and the dalbavancin arm of Seltzer et al.48. Oritavancin doses were considered separately. Six trials with comparator arms where patients continued current therapy or received an option of treatments after randomization were considered too heterogeneous and were excluded from the analysis48–53. Five trials were excluded as TOC or ECR were not reported in the ITT population21^,54–57. A further five trials, all published in 1990 or earlier, investigated treatments that could not be connected to the evidence network and, thus, were not included in the analysis58–62. The Teras et al.63 study was excluded as it was a sub-group analysis of the Breedt et al.64 trial. Summaries of the 33 trials included in the NMA are provided in Table 1.

Table 1. Summary of studies include in network meta-analysis with results for TOC ITT outcomes.

Ref	Authors, year	Treatment	TOC r	ITT n	ECR or sub-groups reported	Key pathogens	Mean age	Male (%)	Mean weight (kg)
68	Aikawa et al., 2013	Daptomycin	NR	55	MRSA	MRSA	–	53.4	–
		Vancomycin	NR	19			–	68.2	–
18	Boucher et al., 2013	Dalbavancin	NR	288	ECR	NR	48.8	59	–
		Vancomycin	NR	285			48.9	60.7	–
64	Breedt et al., 2005	Tigecycline	220	261	MRSA, MSSA	Mixed	48.8	60.9	82.5
		Vancomycin	225	259			50.1	60.6	81.5
69	Corey et al., 2010	Ceftaroline	304	351	MRSA, MSSA	Mixed	47.2	62.7	–
		Vancomycin	297	347			49.2	62.8	–
17	Corey et al., 2013	Oritavancin	416	503	ECR, MRSA	Mixed	45	67.2	76.2
		Vancomycin	404	502			44.4	68.3	78
3	Corey et al., 2014	Oritavancin	378	475	ECR, MRSA, MSSA	Mixed	46.2	63.4	81.9
		Vancomycin	383	479			44.3	62.8	82.7
70	Covington et al., 2011	JNJ-Q2	55	83	ECR, MRSA, MSSA	Mixed	36.9	64.6	–
		Linezolid	48	78			36.9	64.6	–
71	Craft et al., 2011	CEM-102	67	78	ECR, MRSA	S. aureus	41.5	72	–
		Linezolid	73	77			40.6	65	–
72	Fang et al., 2012	Linezolid	288	335	ECR	Mixed	43.3	–	–
		Tedizolid	284	332			43.3	–	–
73	Dunbar et al., 2011	Oritavancin 200 mg	63	87	MRSA, MSSA	Mixed	–	–	–
		Oritavancin 800 mg (with option of additional 400 mg)	68	87			–	–	–
		Oritavancin 1200 mg	72	88			–	–	–
74	Fang, 2013	Linezolid	293	334	–	Mixed	45.6	64.1	–
		Tedizolid	292	332		Mixed	45.6	67.8	–
75	Florescu et al., 2008	Tigecycline	NR	70	MRSA	MRSA	–	65	80.1
		Vancomycin	NR	23			–	59	75.5
16	Itani et al., 2010	Linezolid	NR		MRSA	MRSA	49.7	57	85.4
		Vancomycin	NR				49.4	61	85.9
76	Jauregui et al., 2005	Dalbavancin	505	571	–	Mixed	47	62	–
		Linezolid	246	283			46	61	–
23	Kohno et al., 2007	Linezolid	NR	17	MRSA	MRSA	68.4	70	50.7
		Vancomycin	NR	10			67.5	70.6	53
77	Krievins et al., 2009	Iclaprim 0.8	28	30	MRSA	Mixed	49.7	70	85
		Iclaprim 1.6	29	32			45.6	78.1	84.6
		Vancomycin	28	30			43.5	73.3	82.7
78	Noel et al., 2008	Ceftobiprole	448	547	MRSA, MSSA	Mixed	52.9	63	–
		Vancomycin	227	281			48	55	81
79	Noel et al., 2008	Ceftobiprole	309	397	MRSA, MSSA	Mixed	46.7	61	82.7
		Vancomycin	300	387			51.9	64	–
65	Noel et al., 2012	Omadacycline	98	111	MRSA	Mixed	–	53	–
		Linezolid	82	108			36.2	70.8	–
66	Noel et al., 2012	Linezolid	64	72	–	Mixed	–	59	–
		Omadacycline	58	68			41.7	61.8	–
80	Pertel et al., 2009	Daptomycin	36	50	–	S. aureus	–	34	–
		Vancomycin	28	51			–	49	–
81	Postier et al., 2004	Tigecycline	NR	4	MRSA, MSSA	Mixed	49.3	67	85.8
		Tigecycline	NR	4			48.7	80	86.3
82	Prince et al., 2013	BC-3781 100	NR	66	ECR	Mixed	41.7	68.6	91.3
		BC-3781 150	NR	332			42.2	66.2	96.7
		Vancomycin	NR	335			40.4	59.1	89
19	Prokocimer et al., 2013	Tedizolid	284	332	ECR, MRSA, MSSA	Mixed	43.1	59.1	–
		Linezolid	288	335			43.6	61.4	–
83	Sacchidanand et al., 2005	Tigecycline	209	277	MRSA, MSSA	Mixed	49.4	61.6	81.5
		Vancomycin	200	260			48.4	66.9	82
22	Sharpe et al., 2005	Linezolid	NR	30	MRSA	MRSA	66	33	–
		Vancomycin	NR	30			76	50	–
84	Stevens et al., 2000	Linezolid	279	400	MRSA	Mixed	46.8	63	79.1
		Oxacillin-dicloxacillin	272	419			49.2	60.9	79
49	Stryjewski et al., 2005	Telavancin	66	84	MRSA	Mixed	44.3	55	–
		Standard therapy	66	83			44.6	64	–
50	Stryjewski et al., 2006	Telavancin	82	100	MRSA	Mixed	42.3	65	–
		Standard therapy	81	95			44.7	55	–
85	Stryjewski, 2012	TD-1792	79	98	ECR, MRSA, MSSA	Mixed	40	62	–
		Vancomycin	82	99			40	60	–
86	Talbot et al., 2007	Ceftaroline	59	67	–	Mixed	44	59.4	–
		Vancomycin	26	32			41.6	55.2	–
20	Weigelt et al., 2005	Linezolid	439	476	MRSA, MSSA	Mixed	51.7	63.3	–
		Vancomycin	402	454			51.8	61.7	–
87	Wilcox et al., 2004	Linezolid	99	106	–	S. aureus	53	54	75
		Teicoplanin	89	102			55	54	73
88	Wilcox et al., 2010	Ceftaroline	291	342	MRSA, MSSA	Mixed	47.8	65.5	–
		Vancomycin	289	338			47.5	59.5	–
89	Wilcox, 2013	Dalbavancin	NR	371	ECR	NR	49.1	60.1	–
		Vancomycin	NR	368			51.4	54.6	–

The definition of ECR varied widely across studies; definitions included characterization solely by absence of fever or absence of fever as well as absence of need for rescue antibiotics. The evidence network for ECR included only 10 studies, while TOC included 26 studies.

Results of base case mixed treatment comparisons

In total, there were 19 treatments in 26 studies that reported TOC and were connected to the evidence network (Figure 2). Clindamycin (and combinations) and telavancin did not connect to the evidence network, so relative treatment effects could not be calculated. Treatment effects are presented as odds ratios vs vancomycin, with ratios greater than one indicating a superior TOC outcome for the comparator (Table 2). The posterior mean and 95% credible intervals (CrI) for odds ratios of TOC for antimicrobial agents relative to vancomycin were linezolid (OR = 1.55; 95% CrI = 0.91–2.57), dalbavancin (OR = 1.76; 95% CrI = 0.84–3.61), daptomycin (OR = 2.18; 95% CrI = 0.90–5.42), and oritavancin 1200 mg (OR = 1.06; 95% CrI = 0.80–1.43), showing no evidence of a difference in efficacy. Omadacycline showed a statistically significant improvement in TOC over vancomycin (OR = 2.48; 95% CrI = 1.10–5.43). All other 95% credible intervals crossed one and, therefore, none of the other treatments showed a statistically significant difference in TOC compared to vancomycin.

Figure 2. Evidence network for TOC end-point ITT analysis. The thickness of the lines represents the number of studies available for each comparison and the size of the nodes represents the number of studies available for each treatment.

Table 2. Mean odds ratios (95% credible intervals) of treatments vs vancomycin for base case and sub-group analyses.

Treatment	TOC ITT	ECR ITT	TOC MRSA	TOC MSSA	TOC CE	ECR CE	TOC ITT FDA	TOC CE FDA
	Random effects (DIC = 476.4)*	Random effects (DIC = 137.6)	Random effects (DIC = 436.3)	Random effects (DIC = 243.2)	Random effects (DIC = 346.1)	Fixed effects (DIC = 44.65)	Fixed effects (DIC = 83.40)	Fixed effects (DIC = 146.90)
BC-3781 100	–	1.95 (0.16, 25.51)	–	–	1.60 (0.12, 4.07)	–	–	0.94 (0.17, 2.85)
BC-3781 150	–	1.4 (0.13, 15.44)	–	–	2.74 (0.11, 3.8)	–	–	0.83 (0.16, 2.53)
Ceftaroline	1.07 (0.77, 1.51)	–	0.82 (0.26, 2.61)	0.76 (0.14, 4.18)	2.29 (0.21, 2.48)	–	–	–
Ceftobiprole	1.04 (0.76, 1.42)	–	1.31 (0.47, 3.65)	0.98 (0.17, 5.30)	1.63 (0.31, 3.41)	–	–	–
CEM-102	0.48 (0.11, 1.82)	–	0.36 (<0.01, 27.44)	–	0.65 (0.01, 2.3)	–	–	–
Dalbavancin	1.76 (0.84, 3.61)	1.01 (0.24, 4.41)	–	–	1.16 (0.47, 2.35)	–	–	1.04 (0.55, 1.77)
Daptomycin	2.18 (0.90, 5.42)	–	0.74 (0.13, 3.66)	–	–	–	–	–
Iclaprim 0.8	1.06 (0.11, 12.85)	–	–	–	–	–	–	–
Iclaprim 1.6	0.68 (0.08, 4.88)	–	–	–	–	–	–	–
JNJ-Q2	1.90 (0.79, 4.48)	–	1.41 (0.35, 5.54)	1.55 (0.05, 45.92)	–	–	–	–
Linezolid	1.55 (0.91, 2.57)	–	1.55 (0.96, 2.46)	1.36 (0.15, 13.34)	1.65 (0.68, 3.53)	–	–	–
Omadacycline	2.48 (1.10, 5.43)	–	6 (0.32, 172.43)	–
Oritavancin	1.06 (0.80, 1.43)	1.02 (0.23, 4.33)	0.94 (0.44, 2.02)	0.82 (0.08, 7.83)	1.24 (0.41, 2.26)	0.89 (0.65, 1.16)	1.07 (0.85, 1.33)	1.03 (0.71, 1.45)
Oritavancin 200 mg	0.62 (0.26, 1.43)	–	1.32 (0.24, 7.25)	0.08 (<0.01, 2.70)	2.96 (0.12, 2.68)	–	0.66 (0.29, 1.30)	0.65 (0.26, 1.38)
Oritavancin 800 mg	0.84 (0.34, 1.99)	–	2.58 (0.42, 18.52)	0.15 (<0.01, 5.13)	9.44 (0.16, 3.60)	–	0.91 (0.38, 1.83)	0.87 (0.33, 1.89)
Oxacillin-dicloxacillin	1.24 (0.64, 2.39)	–	–	–	3.21 (0.28, 5.09)	–	–	–
TD-1792	0.87 (0.40, 1.87)	0.99 (0.12, 8.20)	1.71 (0.20, 16.68)	0.65 (0.01, 53.25)	1.79 (0.25, 5.25)	–	0.92 (0.41, 1.79)	1.33 (0.38, 3.45)
Tedizolid	1.51 (0.82, 2.73)	–	1.23 (0.44, 3.30)	0.57 (0.02, 15.39)	–	–	–	–
Teicoplanin	0.72 (0.22, 2.21)	–	–	–	–	–	–	–
Tigecycline	0.87 (0.61, 1.26)	–	0.9 (0.24, 3.28)	0.63 (0.11, 3.60)	0.90 (0.31, 1.79)	–	–	–

*DIC is Deviance Information Criterion, used for comparing fixed and random effects models.

The network for ECR included six treatments in 10 studies so was much more limited than the TOC end-point, allowing fewer indirect comparisons, in particular omitting linezolid (Figure 3). The 95% credible intervals around the mean odds ratio relative to vancomycin of ECR for dalbavancin (OR = 1.01; 95% CrI = 0.24–4.41), oritavancin 1200 mg (OR = 1.02; 95% CrI = 0.23–4.33), and all other treatments crossed unity. There was, therefore, no statistically significant evidence of improvement in ECR compared to vancomycin (Table 2).

Figure 3. Evidence network for ECR end-point ITT analysis. The thickness of the lines represents the number of studies available for each comparison and the size of the nodes represents the number of studies available for each treatment.

Fixed effects results for both these analyses were similar and are presented in the Supplementary Appendix.

Results of subgroup analyses—MRSA and MSSA populations

The evidence networks of the sub-groups for the TOC end-point in the ME population analysis are shown in the Supplementary Appendix. The MRSA network included 15 treatments in 25 studies, while the MSSA network was more limited, with 11 treatments in 13 studies. The ECR outcome was reported in only five studies with MRSA sub-groups and four with MSSA sub-groups, and this evidence was too limited to conduct a meaningful network meta-analysis (Table 1).

For the MRSA sub-group (Figure 4), the odds ratios relative to vancomycin for TOC were linezolid (OR = 1.55; 95% CrI = 0.96–2.46), daptomycin (OR = 0.74; 95% CrI = 0.13–3.66), and oritavancin 1200 mg (OR = 0.94; 95% CrI = 0.44–2.02). In the MSSA sub-group (Figure 5) they were linezolid (OR = 1.36; 95% CrI = 0.15–13.34) and oritavancin 1200 mg (OR = 0.82; 95% CrI = 0.08–7.83). The large uncertainties around some of the results were due to the sparse data available for these treatments. All of the 95% credible intervals for MRSA and MSSA populations crossed unity so none of the treatments led to a statistically significant difference in TOC response compared to vancomycin. The MRSA results for iclaprim 0.8 mg are not presented in the table as they were based on only four MRSA patients in a single study. Fixed effects results are presented in the Supplementary Appendix.

Figure 4. Evidence network for TOC end-point ME population analysis in MRSA sub-group. The thickness of the lines represents the number of studies available for each comparison and the size of the nodes represents the number of studies available for each treatment.

Figure 5. Evidence network for TOC end-point ME population analysis in MSSA sub-group. The thickness of the lines represents the number of studies available for each comparison and the size of the nodes represents the number of studies available for each treatment.

Results of the four sensitivity analyses

The TOC mixed CE population analysis included 14 treatments in 20 studies. All 95% credible intervals included unity, indicating no strong evidence of a difference in efficacy between vancomycin and any of the comparators (Table 2).

Only the SOLO I and SOLO II trials were included in the ECR mixed CE populations analysis3^,17, so only oritavancin 1200 mg and vancomycin were comparable on this end-point and population. Fixed effects results are presented in Table 2, as there were fewer than three studies in the analysis. The posterior mean and 95% credible intervals odds ratio of oritavancin 1200 mg relative to vancomycin was OR = 0.89; 95% CrI = 0.65–1.16, and gave no evidence of difference in efficacy.

The TOC mixed ITT FDA analysis was limited to five treatments in four studies. The limited evidence caused the random effects analysis to be unstable. Fixed effects results are, therefore, presented in Table 2. There was again no evidence of a difference in efficacy for TOC between each of the comparators and vancomycin.

The TOC mixed CE FDA populations analysis was limited to eight treatments in seven studies, so the random effects results were again unstable. Table 2, therefore, presents the fixed effects analysis. There was no evidence of superiority of any treatment over vancomycin.

For all sensitivity analyses, results from the alternative, random, or fixed effects, models are presented in the Supplementary Appendix.

Consistency checking

The evidence networks were inspected to identify loops, not due solely to three arm trials; only one case where inconsistency could be present was identified in the TOC mixed (CE) network. This loop consisted of vancomycin, linezolid, and dalbavancin. The node-splitting method found no substantial differences in the direct and indirect treatment effect estimates and, therefore, no evidence of inconsistency in this network.

Discussion

The objective of this study was to conduct a systematic review and formal comparison to demonstrate similar efficacy of oritavancin to other available antimicrobial agents for the treatment of ABSSSI, due to a mixture of microbial causes, on the basis of clinical and microbiological success at TOC and on the basis of ECR. The systematic review identified 52 trials. The Bayesian NMA provided a framework for the formal comparison of treatments included in the identified trials. This NMA found evidence that oritavancin 1200 mg was equivalent to vancomycin (OR = 1.06; 95% CrI = 0.80–1.43) at TOC. The indirect evidence also suggests that oritavancin 1200 mg demonstrates equivalence to linezolid (OR = 1.55; 95% CrI = 0.91–2.57) and daptomycin (OR = 2.18; 95% CrI = 0.90–5.42) at TOC. The only significant finding was that omadacycline was superior to vancomycin for TOC in the mixed population (OR = 2.48; 95% CrI = 1.10–5.43). However, only two trials (a total of 359 patients) compared omadacycline to linezolid and these studies had contradictory findings65^,66. There were no other statistically significant findings of differences in efficacy between the anti-microbial agents in the mixed populations for TOC or ECR or for TOC in the MRSA or MSSA sub-groups. Sensitivity analyses found the results of the base case to be robust to choice of ITT over CE population and the choice to include studies that did not conform to recent FDA guidance on clinical trials.

This is the first systematic review and NMA to include oritavancin and an assessment of ECR in ABSSSI patients. The results for all outcomes and populations were largely consistent with the head-to-head findings of the included clinical trials. Three previous systematic reviews and meta-analyses found linezolid to be more efficacious for the treatment of ABSSSI than vancomycin, with mean and 95% confidence intervals (CI) for the odds ratios of OR = 1.57; 95% CI = 1.18–2.1028, OR = 1.40; 95% CI = 1.01–1.9531, and OR = 1.67; 95% CI = 1.31–2.1233. However, the present study is based on a larger and more recent set of trials and includes indirect evidence via the NMA methodology. A recent large meta-analysis24 of 53 studies in the treatment of ABSSSI comparing vancomycin with linezolid and other anti-microbial agents found only significant differences in efficacy for linezolid (OR = 1.61; 95% CI = 1.07–2.43). This study also neglected indirect evidence and was restricted to comparisons with vancomycin. These results from earlier analyses are consistent with the mean and 95% credible interval for the odds ratio relative to vancomycin for linezolid (OR = 1.55; 95% CrI = 0.96–2.46) found in the present NMA.

One previous NMA assessed the comparative efficacy of vancomycin, dalbavancin, linezolid, telavancin, daptomycin, and tigecycline for MRSA complicated skin and soft tissue infections7. The search was conducted in 2008 and was, thus, limited to only 14 studies in MRSA, while the present study identified 25 studies in MRSA. Their odds ratios for linezolid (1.83), daptomycin (1.21) and tigecycline (0.81) were comparable to the estimates in the MRSA sub-group of the present study.

Two further meta-analyses have compared daptomycin with non-daptomycin treatments in skin and soft-tissue infections, one due only to MRSA29 and another due to both MRSA and MSSA12. The study in only MRSA caused infections found a mean and 95% confidence interval for the odds ratio of daptomycin relative to other treatments of OR = 0.89; 95% CI = 0.63–1.2529, consistent with the odds ratio for daptomycin relative to vancomycin in the MRSA sub-group of the present study (OR = 0.74; 95% CrI = 0.13–3.66). The study in ABSSSI due to a mixture of MRSA and MSSA found a mean and 95% confidence interval for the odds ratio of daptomycin relative to vancomycin of OR = 1.05; 95% CI = 0.84–1.31) for clinical success12, consistent with the odds ratio for TOC of daptomycin relative vancomycin in ABSSSI in the present study (OR = 2.18; 95% CrI = 0.90–5.42).

Limitations

Across all outcomes and populations, wide credible intervals were seen, pointing to the heterogeneity of available evidence. For the ECR end-point, it was only possible to indirectly compare five treatments to the common comparator, vancomycin. The systematic review was limited to English language publications; it is possible non-English language trials have been missed.

As recommended by FDA guidance, the base case analyses for the TOC and ECR end-points were conducted in the ITT populations. Using the ITT population, when a high numbers of dropouts across the studies are observed, could be considered a potential limitation. The impact of this was assessed in sensitivity analyses including patients in the CE populations, i.e. excluding patients who dropped out of the studies or were otherwise not evaluable; results were similar to the base case. Differences in the definition of ABSSSI and other patient characteristics across trials were a further limitation. Sensitivity analyses restricted to studies conducted in accordance with recent FDA guidelines5, thus ensuring greater similarity in populations, gave results similar to the base case.

The NMA methodology suffers from several limitations due to the assumptions necessary to conduct the indirect comparisons39^,67. Fixed effects models must assume that the relative treatment effect across studies is the same and this may not be the case due to heterogeneity38^,40. Random effects models weaken this to an assumption of exchangeability of treatment, but these models can be difficult to fit or lead to highly uncertain estimates. NMA makes a further assumption of consistency of treatment effects across studies43, although no evidence of inconsistency in the networks was identified45.

Further research

As discussed, wide credible intervals were seen across the results of these analyses and several important comparisons in TOC, ECR, and in the sub-groups of MRSA and MSSA infections were not possible due to limitations in the evidence base. A large and high quality trial directly comparing several antimicrobial agents could be of high value to clinical practice and healthcare decision-making.

Conclusion

This systematic review and NMA has provided an up-to-date and unified framework for the comparison of available antimicrobial agents used in the treatment of ABSSSI; this is the first such analyses to include novel agents such as oritavancin and to look at the ECR end-point. The results suggest equivalence between oritavancin 1200 mg, and other novel antimicrobial agents including daptomycin and linezolid, compared with vancomycin for both TOC and ECR.

Transparency

Declaration of funding

The Medicines Company funded ICON plc to complete this research.

Declaration of financial/other relationships

HT was a paid consultant of ICON Plc for the duration of this project. DAS, JCT and NH are employees of ICON Plc which was funded by The Medicines Company to complete this research. KS is an employee of The Medicines Company. GRC is an advisor/consultant to The Medicines Company, Pfizer, Merck, GSK, Cubist and others. CMRO peer reviewers on this manuscript have received an honorarium from CMRO for their review work, but have no other relevant financial relationships to disclose.