AntiBiotic Choice and Duration (ABCD) Trial
Heller Chen, Tim Jenkins, Aline Maddux, and Christiana Smith
1. Background and Setting
Skin and soft tissue infections are among the most common bacterial infections, leading to 14 million health care visits each year in the United States1. Skin and soft tissue infections also result in nearly 900,000 hospitalizations each year2, and the rate of hospitalizations increased 71% between 1997 and 20051. The main types of skin and soft tissue infections are cellulitis, cutaneous abscesses, and infected ulcers; the most common of these leading to hospitalization is cellulitis, a diffuse, spreading infection of the dermis and subcutaneous tissue. Despite the frequency of cellulitis, there have been few randomized trials of the treatment of patients with infection severe enough to require hospitalization; therefore, the optimal treatment in this clinical setting remains unclear.
What is the necessary spectrum of antimicrobial activity?
Studies in the 1980’s and 1990’s demonstrated that the majority of cases of cellulitis were caused by beta-hemolytic streptococci, particularly Group A streptococcus, and less commonly methicillin-susceptible Staphylococcus aureus3,4 However, the emergence of community-associated methicillin-resistant S. aureus (CA-MRSA) in the late 1990’s led to a marked change in the epidemiology and microbiology of skin infections in the United States; the majority of purulent skin infections (most frequently abscesses) are now caused by MRSA5. However, because the infecting pathogen in cases of cellulitis is usually not able to be determined, it has been unclear whether CA-MRSA has also become a common cause of cellulitis.
Several observational studies since the start of the CA-MRSA epidemic have suggested that MRSA is an infrequent cause of cellulitis based on clinical response to antibiotic therapy and/or microbiologic findings. For example, in studies of outpatient adults and children, treatment of cellulitis with antibiotics with MRSA activity did not lead to improved cure rates compared with beta-lactam antibiotics (which have no activity against MRSA)6,7. In addition, the MRSA-active antibiotics in these studies were associated with more frequent adverse drug events than beta-lactams. In another study of adults hospitalized with cellulitis, detailed serologic and microbiologic studies demonstrated that 131 of 179 (73%) cases were caused by beta-hemolytic streptococci, and 92 of 96 (96%) patients treated with a beta-lactam antibiotic alone responded to therapy8. The only randomized trial to address this question evaluated outpatients with cellulitis presenting to an emergency department9. In this study, all participants received oral cephalexin (a beta-lactam without MRSA activity) and were randomized to the addition of trimethoprim-sulfamethoxazole (an MRSA-active agent) versus placebo. Clinical outcomes were similar between the two groups, suggesting that the addition of an MRSA-active agent to a beta-lactam does not improve clinical outcomes.
In contrast, one single-center observational study of patients treated for cellulitis in an outpatient clinic in Hawaii demonstrated that treatment success rates were lower with beta-lactam antibiotics compared with MRSA-active agents10. This finding might be attributable to increased rates of MRSA carriage in different populations or geographic locations, with some studies suggesting a disproportionate burden of MRSA disease among the Pacific Islander population of Hawaii11. Despite this conflicting evidence, the Infectious Diseases Society of America (IDSA) guideline for the management of skin and soft tissue infections recommends beta-lactam antibiotic therapy for patients hospitalized with cellulitis12. In clinical practice, however, the majority of patients hospitalized with cellulitis are treated with MRSA-active antibiotics, most commonly vancomycin. In a large multicenter observational study, less than 20% of inpatients with cellulitis were treated with beta-lactams, while over 70% were treated with vancomycin13. This suggests that in the era of CA-MRSA, in patients with cellulitis severe enough to warrant hospitalization, most providers consider MRSA to be a relevant pathogen. Since MRSA-active antibiotics may be associated with more adverse events and higher costs than beta-lactams, rigorous data are needed in patients hospitalized with cellulitis to determine the necessary spectrum of antimicrobial activity.
What is the shortest effective duration of therapy?
The optimal duration of antibiotic therapy for patients hospitalized with cellulitis is not known. The limited available evidence suggests that shorter treatment courses may be as effective as longer courses. In one small randomized trial, 5 days of levofloxacin was as effective as 10 days; however, only 10% of patients in this study required hospitalization, and only patients who were clinically improving by day 5 of treatment were randomized, thus limiting the generalizability of the study14. In a recent industry-sponsored randomized trial in adults with either cellulitis, abscess, or wound infection, 6 days of tedizolid was as effective as 10 days of linezolid15. However, it is not known whether these findings are antibiotic- or class-specific and can be extrapolated to antibiotics more commonly used to treat cellulitis.
Based on the currently available evidence, the IDSA guideline recommends 5 days of therapy for cellulitis, with consideration of extension of the treatment duration for cases that have not clinically responded12. However, in clinical practice, most patients hospitalized with cellulitis are treated substantially longer than 5 days; in a multicenter cohort, the median duration of therapy was 12 days (interquartile range 10 – 15 days), and fewer than 10% were treated for 7 days or less13.
Overall, these findings highlight that the evidence for the appropriate duration of therapy for patients hospitalized with cellulitis is insufficient. A randomized trial delineating the shortest effective treatment duration in patients hospitalized with cellulitis is necessary to inform the most appropriate clinical practice. If shorter durations are found to be effective in this clinical setting, antibiotic exposure could be substantially reduced, thereby reducing unintended consequences of antibiotics such as the development of resistance and adverse events, as well as reducing costs associated with excessive antibiotic use and prolonged hospitalization.
Proposed clinical trial
In order to fill the evidence gaps discussed above, we propose the following multi-center, Phase III clinical trial. The purpose of the trial is to compare the effect of MRSA-active versus non-MRSA-active antibiotic therapy and 5-day versus 10-day treatment courses on the probability of treatment success in adults hospitalized with cellulitis. We hypothesize that: 1) non-MRSA active antibiotics lead to a non-inferior treatment success rate compared with MRSA-active antibiotics and 2) 5-day treatment courses lead to a non-inferior treatment success rate compared with 10-day treatment courses.
2. Synopsis of Proposed Study
Study population and setting
We will conduct a double-blind, randomized, placebo-controlled trial with a 2 x 2 factorial and non-inferiority design to determine the effect of antibiotic choice and duration on the probability of treatment success in adults hospitalized with cellulitis. We will include patients who are admitted to the hospital with a primary diagnosis of non-purulent cellulitis. We will exclude patients who are less than 18-years-old, immune suppressed (neutropenic, on active chemotherapy or steroids or have received a bone marrow transplant), those transferred from an outside hospital after admission for longer than 24 hours, those with infection due to vascular insufficiency, allergy to any antibiotics used in the trial, antibiotic use within two weeks of presentation, pregnant women, and those with bacteremia, severe sepsis or septic shock, or deep tissue infection (abscess, fasciitis, myositis, osteoarthritis, or septic arthritis).
Study treatments
All enrolled subjects will be initially treated with 5 days of MRSA-active or non-MRSA-active antibiotic therapy (Figure 1). We will use cefazolin (intravenous) and cephalexin (oral) as the non-MRSA-active antibiotics and vancomycin (intravenous) and clindamycin (oral) as the MRSA-active antibiotics. For the duration of the inpatient stay, the intravenous agent will be administered. At the time of hospital discharge, therapy will be transitioned to the appropriate oral agent; the average time until discharge is anticipated to be 3 to 4 days. After the initial 5 days of antibiotic therapy, subjects randomized to the 10-day treatment duration will receive an additional 5 days of antibiotic therapy, whereas subjects randomized to the 5-day treatment duration will receive 5 days of placebo (Figure 1).
Figure 1.
Study endpoints
The primary endpoint will be measured 7-10 days after completion of the 10-day course of study medications (on study day 17-20) (Figure 1). The primary endpoint will be treatment success, defined as lack of treatment failure. Treatment failure is defined as change, extension, or re-initiationof antibiotic therapy due to lack of symptom resolution or worsening of symptoms, medication side effects resulting in discontinuation or change in antibiotic, infection at a new body site, unanticipated surgical intervention, or re-hospitalization related to infection. We will assess secondary endpoints, including individual components of treatment failure, serious adverse events, and a secondary assessment of treatment success at study day 30.
Study measurements
At enrollment, all subjects will be interviewed regarding their medical history and home medications (see study visit table, Appendix I). Vital signs (oral or axillary temperature, heart rate, blood pressure, respiratory rate) and a physical examination targeted to the area of cellulitis will be performed. All physical examinations will be performed by site investigators and will include the surface area of erythema (greatest head-to-toe height multiplied by greatest side-to-side width), degree of warmth, erythema, induration, and tenderness, and presence of lymphadenopathy, lymphangitis, drainage, ulceration, or bullae. After the enrollment examination, subjects deemed to have a skin infection other than non-purulent cellulitis or a non-infectious skin condition will be withdrawn. Laboratory and microbiology tests including a complete blood count with differential, comprehensive metabolic panel, and blood cultures will be obtained upon enrollment. A serum vancomycin trough level will be obtained prior to the 4th dose of antibiotic. Each day until hospital discharge and at the end of treatment visit, vital signs, physical examination, documentation of all study and non-study antibiotics administered, and assessment of adherence to study drug, and adverse events will be performed. At the early and late test of cure visits, the site investigator will assess the primary and secondary endpoints and again evaluate foradverse events. All data will be collected on standardized case report forms.
Statistical Design: Sample size considerations
We will use a binomial probability model and functional outcomes to represent the proportion of subjects with treatment success in each group (θ5, θ10, θN, θY; Table 1). The contrast outcomes will be calculated using the risk difference. The effect of MRSA coveragewill be calculated by:
θN-θY= ((θ5N+θ10N)/2) – ((θ5Y+θ10Y)/2).
The effect of antibiotic duration will be calculated by:
Table 1. Functional Outcomes / No MRSA coverage (cefazolin/cephalexin) / MRSA coverage(vancomycin/clindamycin)
5 days antibiotic duration / θ5N / θ5Y
10 days antibiotic duration / θ10N / θ10Y
θ5- θ10= ((θ5N+θ5Y)/2) – ((θ10N+θ10Y)/2).
Sample size was calculated by:
((zα + zβ)/(θ+ - θ-))2 x (θ(1-θ)) = n/group of pairwise comparison.
We selected a non-inferiority margin of -0.1 based on IDSA recommendations12. We anticipate a baseline 80% success rate based on previous studies10,15. We will assume there is no interaction between antibiotic choice and duration. With a power of 90% and alpha 0.05, we will require atotal sample size of 336 (84 per group) to detect the specified differences in each pairwise comparison. We expect a 15% loss to follow-up rate. Therefore, we plan to enroll a total of 396 subjects (99 per group) to ensure an adequate number for analysis (Table 2).
Table 2. Sample size / No MRSA coverage / MRSA coverage / Total5 days / 99 / 99 / 198
10 days / 99 / 99 / 198
Total / 198 / 198 / 396
Potential inference: With 168 subjects per group the standard error is given by se2 = 2x0.16/168 = 0.0442 therefore the lower limit of the 95% CI is greater than -0.1 if the observed effect is greater than -0.1 + 1.96se = -0.014; i.e., a 1.4% decrease in success rate with shorter duration (or non-MRSA) therapy than with its comparator.
Statistical Design: Analysis plan
All outcomes will be analyzed using an intention-to-treat model. As a secondary analysis, the differences in mean rate of treatment success will be analyzed in the per-protocol population after excluding subjects who withdrew, were lost to follow-up, did not complete the assigned study treatment, or received non-study antibiotics.
We will assess for an interaction effect between antibiotic choice and duration. If there is no interaction, the primary analysis will be performed using a 2-sample t-test of the differences in mean treatment success rates between the MRSA-active and non-MRSA-active treatment groups, and between the 5 days vs. 10 days antibiotic duration groups. Several possible outcomes of this trial are depicted in Figure 2.
Figure 2. Four possible outcomes of the trial are shown. Each scenario assumes no interaction between antibiotic choice and duration. Scenario 1 represents the hypothesis: that non-MRSA-active antibiotics are noninferior to MRSA-active antibiotics, and 5 days of antibiotics is noninferior to 10 days of antibiotics. In scenario 1, 5 days of non-MRSA-active antibiotics would be recommended. In scenario 2, non-MRSA-active antibiotics are inferior to MRSA-active antibiotics, but 5 days of antibiotics is noninferior to 10 days. In scenario 2, 5 days of MRSA-active antibiotics would be recommended. In scenario 3, non-MRSA-active antibiotics are inferior to MRSA-active antibiotics, and 5 days of antibiotics is inferior to 10 days. In scenario 3, 10 days of MRSA-active antibiotics would be recommended. In scenario 4, non-MRSA active antibiotics are noninferior to MRSA-active antibiotics, but 5 days of antibiotics is inferior to 10 days. In scenario 4, 10 days of non-MRSA-active antibiotics would be recommended.
Each of the 4 scenarios shown in Figure 2 assumes there is no interaction effect between antibiotic choice and treatment duration. If an interaction is found, 1-way ANOVA will be used to determine which one of the four treatment groups had the highest treatment success rate.
Interpretation of results
If our hypotheses are correct and the non-MRSA-active therapy is found to be non-inferior to MRSA-active therapy, then non-MRSA-active therapy will be recommended for routine practice given the narrower spectrum of activity and lower incidence of adverse drug events. If the 5-day antibiotic duration is found to be non-inferior to 10 days, then 5-day courses will be recommended for routine practice given the lower risk for antibiotic-related complications.
3. Study Implementation and Conduct
Subject recruitment and retention
We will recruit study subjects from 5 predetermined hospitals (sites). These sites will represent a geographically, racially/ethnically, and socioeconomically diverse patient population. We anticipate enrollmentof approximately sixsubjects per month per site; therefore, we expect study enrollment duration of 13.5 months. At each site, we will address emails to all hospitalists and emergency room physicians informing them about the study, and provide contact information so that study coordinators can be notified of potential subjects. In addition, we will post flyers reminding providers about the study in physician workrooms in the emergency department and on inpatient floors. Physicians will notify the study coordinator when a patient will be hospitalized due to cellulitis. Each potential subject will be evaluated for eligibility by a study investigator and enrolled if all criteria are met and the patient provides informed consent. Compensation will be provided to subjects for completion of study visits to promote retention.
Randomization
After enrollment, the subject will be randomly assigned to a treatment allocation by a study pharmacist according to predefined randomization tables (Appendix 2), where 5N = 5 days of antibiotics, no MRSA coverage; 5Y = 5 days of antibiotics, MRSA coverage; 10N = 10 days of antibiotics, no MRSA coverage; 10Y = 10 days of antibiotics, MRSA coverage. Subjects will be randomized in blocks of 8, stratified by site, with capability to enroll up to 104 subjects per site. No more than 104 subjects will be enrolled per site, so as to ensure enrollment from a variety of sites to account for geographic variation.
Blinding
A study pharmacist will be available on-site at each location and will be responsible for assigning randomization to a treatment group, preparation of the study drug and delivery of study drug to the inpatient unit. The study pharmacist will have no direct contact with patients, nor will he/she have any role in downstream data collection or analysis.
Intravenous (IV) medications (vancomycin, cefazolin, and placebo) will be delivered in opaque brown IV bags labeled “ABCD study medication” with no other information identifying its contents. IV placebo will contain 0.9% sodium chloride (“normal saline”). Oral medications (clindamycin, cephalexin, and placebo) will be delivered as gel caplets in medication bottles labeled “ABCD study medication” with no other information identifying its contents. The oral placebo will contain an inert substance. Subjects randomized to receive 5 days of antibiotics will be transitioned to placebo on day 6. The providers caring for the patients will decide optimal timing of transition from IV to oral medication. Thus, if a subject randomized to receive 5 days of antibiotics is thought to need IV study drug on day 6, the subject will receive IV placebo until such time that providers decide to transition to oral study drug. In this fashion, we will maintain blinding of patients, physicians, nurses, and other caretakers. If at any time providers raise concerns that a subject has demonstrated clinical worsening, or that continuation of blinding poses serious safety concerns, emergency unblinding will occur and a treatment failure will be recorded. After unblinding, providers will be allowed to treat the subject with an antibiotic that they deem appropriate. Data will continue to be collected on subjects whose treatment allocation was unblinded, to allow for downstream assessment of reasons for treatment failure.
Because vancomycin dosing is dependent upon achieving goal serum drug levels, all subjects will undergo a blood draw for a vancomycin level prior to their 4th dose of IV antibiotics. Only the on-site study pharmacist will have access to the results of the vancomycin levels. For patients randomized to receive vancomycin, the pharmacist will use the vancomycin level to adjust the amount of vancomycin in subsequent doses.
While inpatient, subjects will be visited daily by a study investigator who is blinded to their treatment allocation, and pertinent information will be collected and documented on standardized case report forms, as described in the schedule of study visits (Appendix 1). If discharge occurs prior to day 10, the on-site study pharmacist will prepare the remaining oral study drug doses to be taken at home. After discharge, subjects will be asked to return to an outpatient clinic for additional data collection on day 10 +/-1 (if discharged before day 10), day 17-20, and day 30 +/-3. Blinding will be maintained until all analyses are complete.