AAC Accepted Manuscript Posted Online 18 May 2015 Antimicrob. Agents Chemother. doi:10.1128/AAC.00679-15 Copyright © 2015, American Society for Microbiology. All Rights Reserved. 1 Antibacterial treatment effect estimation in acute bacterial skin and skin structure infection: 2 systematic review and meta-analysis. 3 4 D 5 Jordan E. Cates,a Fanny S. Mitrani-Gold,b,# Gang Li,c,* Linda M. Mundyc o w n 6 lo a d 7 University of North Carolina, Chapel Hill, NCa; GlaxoSmithKline (GSK), Research Triangle e d f 8 Park, NCb; GSK, Collegeville, PAc. ro m h 9 t t p : 10 //a a c 11 .a s m 12 . o r g 13 Running title: Antibacterial treatment effect estimation in ABSSSI / o n 14 N o v 15 e m b 16 e r 1 17 # Address correspondence to [email protected] 5, 2 0 18 1 8 b 19 * Present address: Gang Li, Pfizer, Collegeville, PA y g u 20 e s t 21 22 23 1 24 ABSTRACT 25 A systematic literature review and meta-analysis were conducted to estimate the antibacterial 26 treatment effect for linezolid and ceftaroline to inform on the design of acute bacterial skin and 27 skin structure infection (ABSSSI) non-inferiority trials. The primary endpoints included an early D 28 clinical treatment response (ECTR) defined as cessation of lesion spread at 48-72 hours post- o w n 29 randomization and the test of cure (TOC) response defined as total resolution of the infection at lo a d 30 7-14 days post-treatment. The systematic review identified no placebo-controlled trials in e d f r 31 ABSSSI, four placebo-controlled trials in uncomplicated skin and soft tissue infection as a proxy o m h 32 for placebo in ABSSSI, 12 linezolid ABSSSI trials, three ceftaroline ABSSSI trials, and two t t p : 33 trials for non-antibacterial treatment. The ECTR rates at 48-72 hours, and corresponding 95% //a a c 34 confidence intervals (CI), were 78.7% (95% CI 61.1, 96.3%) for linezolid, 74.0% (95% CI .a s m 35 69.7,78.3%) for ceftaroline, and 59.0% (95% CI 52.8, 65.3%) for non-antibacterial treatment. . o r g 36 Early clinical treatment effect could not be estimated, given no available placebo or proxy for / o n 37 placebo data for this endpoint. Clinical, methodological, and statistical heterogeneity influenced N o v 38 the selection of trials for the meta-analysis of the TOC treatment effect estimation. The pooled e m b 39 estimates of TOC treatment response were 31.0% (95% CI 6.2, 55.9%) for the proxy for placebo, e r 1 40 88.1% (95% CI 81.0, 95.1%) for linezolid, and 86.1% (95% CI 83.7, 88.6%) for ceftaroline. The 5, 2 0 41 TOC clinical treatment effect estimation was 25.1% for linezolid and 27.8% for ceftaroline. The 1 8 b 42 antibacterial treatment effect estimation at TOC will inform on the design and analysis of future y g u 43 non-inferiority ABSSSI clinical trials. e s t 44 Abstract word count: 249 2 45 Key words: acute bacterial skin and skin structure infection (ABSSSI), complicated skin and soft 46 tissue infection (cSSTI), complicated skin and skin structure infection (cSSSI), treatment effect, 47 abscess, skin infection D o w n lo a d e d f r o m h t t p : / / a a c . a s m . o r g / o n N o v e m b e r 1 5 , 2 0 1 8 b y g u e s t 3 48 Introduction 49 Over the past decade, robust clinical, scientific, and regulatory debate has ensued for initiatives 50 to improve the design, execution, and analysis of antibacterial clinical trials (1-4). New trials for 51 acute bacterial skin and skin structure infections (ABSSSI), previously referred to as complicated D 52 skin and skin structure infections (cSSTI), remain important given the rise in incidence of o w n 53 methicillin-resistant Staphylococcus aureus infections and reports of treatment failure (5-8). Per lo a d 54 guidance from the United States (US) Food and Drug Administration (FDA), patient eligibility e d f r 55 for enrollment in ABSSSI trials should be restricted to those with erysipelas, cellulitis, major o m h 56 cutaneous abscesses, and wound infections having a minimal lesion surface area involvement of t t p : 57 75 cm2 (9). For trial endpoints, the traditional test of cure (TOC) endpoint, with the treatment //a a c 58 success defined as total resolution of the infection at 7-14 days post-treatment, remains aligned .a s m 59 with European regulatory guidance, yet the treatment success for the primary efficacy endpoint . o r g 60 aligned with the FDA guidance is defined as cessation of lesion spread after 48-72 hours of / o n 61 treatment (9-11). Revisions to the enrollment and endpoint criteria in recent regulatory guidance N o v 62 for ABSSSI trials necessitate re-evaluation of the antibacterial treatment effect estimation e m b 63 calculated from across-trials comparison of existing data for non-inferiority trial design (4, 12- e r 1 64 16). 5, 2 0 65 1 8 b 66 To inform on future non-inferiority ABSSSI trial design, we conducted a systematic review and y g u 67 meta-analysis of antibacterial treatment effect estimation. Linezolid and ceftaroline were a e s t 68 priori selected as drugs representative of potential active comparators for hospitalized adults 69 with ABSSSI in a global phase 3 clinical development program. The trial data extracted for the 70 systematic review were aligned with regulatory guidance for the enrollment and endpoint criteria 4 71 in ABSSSI. A pre-defined meta-analysis plan defined the efficacy variables, primary endpoints 72 of interest, and computational methods for antibacterial treatment effect estimation in ABSSSI 73 historical trials for non-inferiority margin justification. 74 D 75 Materials and Methods o w n 76 Study design. The systematic review and meta-analysis were designed and executed in lo a d 77 compliance with the Preferred Reporting Items for Systematic Reviews and Meta-analysis e d f r 78 (PRISMA) (17-19). Two independent reviewers conducted computer-based literature searches o m h 79 and a systematic review using the MEDLINE search engine (PubMed, US National Library of t t p : 80 Medicine, National Institute of Health; http://www.ncbi.nlm.nih.gov). Ideally, treatment effect //a a c 81 estimation for a non-inferiority antibacterial trial should be calculated from placebo-controlled .a s m 82 trials to allow for within trial comparison of the antibacterial treatment response as compared to . o r g 83 placebo treatment response (11, 14). Given prior report that placebo-controlled trials were non- / o n 84 existent for complicated skin infections (20), we partitioned the literature searches to identify N o v 85 placebo-controlled trials of any antibacterial treatment for ABSSSI (search A1) as well as e m b 86 randomized clinical trials of active comparators in ABSSSI for linezolid (search B1) and e r 1 87 ceftaroline (search B2). Due to nomenclature changes over time, studies were included that 5, 2 0 88 assessed ABSSSI, complicated skin and soft tissue infections (cSSTI), or complicated skin and 1 8 b 89 skin structure infections (cSSSI). While it was previously reported that there were no historical y g u 90 data for placebo-controlled trials of any antibacterial treatment in complicated skin infections, e s t 91 our search strategy, by design, would have captured any placebo-controlled ABSSSI trials, 92 inclusive of trials for linezolid and ceftaroline. The placebo search was extended to include 93 clinical trials reporting a placebo treatment response in uncomplicated skin and soft tissue 5 94 infections (uSSTI) as a proxy for placebo in ABSSSI (search A2). This strategy, which 95 comprised using a proxy for placebo effect estimation, has been recommended in FDA guidance 96 for non-inferiority trials and has been used for placebo treatment effect estimation for trials in 97 complicated urinary tract infection (13, 21, 22). Given that historical studies published in 1937 D 98 reported relatively high response rates in treatment of erysipelas with ultraviolet (UV) light, an o w n 99 additional search (search C) was conducted to identify clinical trials of non-antibacterial lo a d 100 treatment in uSSTI and in ABSSSI (4, 23, 24). This search was a priori conducted to identify e d f r 101 potentially informative data and not as a proxy for placebo treatment response estimation. o m h 102 t t p : 103 Identification of trials and full-screen review for eligibility //a a c 104 For each search, the publications were restricted to dates before 28 February 2013. Search terms .a s m 105 included abbreviations, plural and full phrase versions of the selected search terms (listed in . o r g 106 Appendix 1) combined with Boolean operators (AND, OR) used in succession to narrow or / o n 107 widen the respective searches. Two reviewers independently assessed the identified publications N o v 108 for trial eligibility, data quality, and efficacy data. Eligible studies included randomized clinical e m b 109 trials for the respective searches of placebo-controlled trials (search A1), active comparator trials e r 1 110 of linezolid (search B1) and ceftaroline (search B2) in ABSSSI, and trials of non-antibacterial 5, 2 0 111 treatment in uSSTI and ABSSSI (search C). Case series, observational studies, review papers, 1 8 b 112 duplicate studies, and studies with incomplete data were not eligible for full text review. y g u 113 Publications identified in searches A1, B1, B2, and C were excluded for one or more of the e s t 114 following criteria: pediatric exclusivity, lack of reported measure of endpoints of interest, more 115 than 15% of subjects with either diabetic foot infections, animal or human bites, necrotizing 116 fasciitis, decubitus ulcer infection, myonecrosis, or ecthyma gangrenosum, and pooled treatment 6 117 response rates for multiple infection types. Publications of uSSTI in search A2 were eligible for 118 full review if the report was a placebo-controlled trial in impetigo, furunculosis, carbunculosis, 119 folliculitis, ecthyma, erysipelas, or secondarily-infected traumatic lesions (SITL). Publications 120 of uSSTI studies were excluded for one of the following criteria: placebo add-on treatment, D 121 pediatric-exclusive studies, prevention or colonization trials, or inclusion of subjects with simple o w n 122 abscesses treated via incision and drainage as surgical treatment in minor cutaneous abscesses. lo a d 123 The latter group was excluded as any potential estimate of placebo effect could not be e d f r 124 independently assessed of the surgical treatment effect (25-27). o m h 125 t t p : 126 Primary endpoints //a a c 127 For the meta-analysis, the two primary endpoints of interest were in the intention to treat (ITT) .a s m 128 population and defined as the early clinical treatment response at 48-72 hours after . o r g 129 randomization and the TOC clinical treatment response. Outcome definitions of clinical success, / o n 130 cure, or treatment failure were evaluated from each of the published studies on the basis of N o v 131 criteria available and assessed for methodological and clinical heterogeneity. Early clinical e m b 132 treatment response was defined as the proportion of subjects who experienced cessation of e r 1 133 spread or the reduction in the total surface area of the lesion at 48-72 hours after randomization 5, 2 0 134 or after the first dose of study drug (9). The TOC clinical treatment response varied in definition 1 8 b 135 based on location and year of the trial execution, but was typically defined as the proportion of y g u 136 subjects who experienced total resolution of all signs and symptoms of the infection at 1-2 weeks e s t 137 after completion of therapy (9). Studies that incorporated improvement in the definition of 138 success were included in the meta-analysis, yet noted for this composite definition of success. 139 Treatment was divided into three broad categories: placebo (with extension to proxy for 7 140 placebo), antibacterial and non- antibacterial treatment for the ITT population unless otherwise 141 noted. 142 143 Data extraction D o 144 Data were systematically extracted for general study characteristics which included author, w n lo 145 publication date, study drug and duration of therapy, active comparator, study population a d e d 146 description, study years, countries or regions, study centers, and outcomes. Baseline f r o m 147 demographic data were abstracted to assess age, gender, geographic region, proportion of h t t 148 baseline infection types and lesions size when available. Methodological details were abstracted p : / / a 149 to assess randomization procedures, sequence generation, blinding in subjects and in a c . a 150 investigators, eligibility criteria, inclusion and exclusion criteria, subject withdrawals, allocation s m . o 151 concealment, primary outcome variable, secondary outcome variable, study definitions, timing of r g / 152 outcome assessment (including early clinical treatment response at 48-72 hours after on N 153 randomization and TOC clinical treatment response), incomplete reporting of outcomes, and o v e m 154 other potential sources of bias. b e r 155 1 5 , 156 Assessments of trial heterogeneity and publication bias 2 0 1 157 Publications reporting trials meeting inclusion criteria were assessed for clinical, methodological, 8 b y 158 and statistical heterogeneity. Clinical heterogeneity was defined by patient selection, g u e 159 interventions, and outcomes; methodological heterogeneity was defined by study design and st 160 execution; statistical heterogeneity was defined as variation in the results beyond sampling 161 variability (28-31). Given the qualitative decisions associated with clinical and methodological 162 heterogeneity, each trial meeting the eligibility criteria was systematically assessed by each 8 163 reviewer for inclusion and exclusion criteria, types and distribution of skin infections, timing of 164 the outcome assessments, treatment duration, inclusion of lesion improvement as an indicator of 165 success, and the proportion of subjects with MRSA pathogens. Each of two independent 166 reviewers employed the Cochrane Risk of Bias Tool for assessment of publication bias and any D 167 potential discrepancies were adjudicated via consensus with a third assessor (32). o w n 168 lo a d 169 Study selection for meta-analysis e d f 170 Extensive clinical, methodological, and statistical heterogeneity influenced the selection of ro m 171 studies included in the meta-analysis (31). The proxy for placebo trials selected for inclusion in h t t p 172 the meta-analysis were restricted to studies that did not exclusively enroll subjects with SITL, :/ / a a 173 given differential regulatory approval of antibacterial agents for this infection type. For the c. a s m 174 linezolid and ceftaroline trials, selection for inclusion in the meta-analysis was based on trial . o r 175 parameters in the historical data that approximated recent ABSSSI guidance, to minimize clinical g / o n 176 and methodological heterogeneity of studies pooled in the meta-analysis (9, 31). These selection N o 177 criteria were trials with total surface lesion area >75 cm2, a non-MRSA specific population, and ve m b 178 abscesses in <50% of the population. The trials reporting UV light treatment response were e r 1 179 included in the meta-analysis as pooled non-antibacterial treatment (20,23,24). 5 , 2 180 0 1 8 181 Data analysis b y g 182 Pooled treatment response rates and 95% confidence intervals (CI) were calculated for each u e s 183 treatment group using the DerSimonian-Laird methodology; a non-iterative, random effects t 184 model to account for inter-study variability utilized the metaphor package in R software (33). 185 This method is commonly employed in meta-analyses of clinical trials, as it accounts for the 186 heterogeneity of studies through a statistical parameter that represents the inter-study variation of 9 187 the trials in the model (33). The proportion of total variation in the study estimates that was due 188 to inter-study statistical heterogeneity was assessed using I2 values and associated p-values for 189 trials included in the meta-analysis. Given the small number of proxy for placebo and active 190 comparator trials, meta-regression was not feasible. The indirect, across-trial comparisons to D 191 estimate the antibacterial treatment effect were calculated as the difference between the lower o w n 192 bound of the 95% confidence interval (CI) of the response rate estimate for each antibacterial lo a d 193 treatment and the upper bound of the of the 95% CI of the proxy for placebo response rate e d f 194 estimate (14,21). This methodology is aligned with FDA guidance on the design of non- ro m h 195 inferiority trials and is acknowledged for yield of intrinsically conservative non-inferiority t t p : 196 margin estimates (14). Publication bias and heterogeneity were assessed using funnel plots (34). //a a c 197 .a s m 198 Results . o r g 199 Systematic Review / o n 200 No placebo-controlled trials of ABSSSI were identified (search A1). There were 800 placebo N o v 201 records for uSSTI (search A2), 477 linezolid records (search B1), 73 ceftaroline records (search e m b 202 B2), and 2390 records for trials defined as non-antibacterial treatments for skin infection (search e r 1 203 C). The final selection of publications included four placebo-controlled uSSTI trials (35-38), 12 5, 2 0 204 linezolid trials (39-50), three ceftaroline trials (51-53), one retrospective ceftaroline secondary 1 8 b 205 analysis (54), and two non-antibacterial treatment trials of UV light for erysipelas (23,24) y g u 206 (Figure 1). e s t 207 208 Proxy trials of uSSTI for placebo treatment of ABSSI (searches A1 and A2). The four trials as 209 proxy for placebo included 234 subjects with a range of uncomplicated infections reported as 10