Consecutive patients attending the participating EDs with cellulitis, who meet the inclusion criteria and none of the exclusion criteria for the trial, will be considered eligible for enrolment. Eligibility for enrolment will be determined by a suitably trained member of the research team.
Inclusion criteria
-
1.
Clinically diagnosed cellulitis, affecting any body part, excluding the perineum, and having any two of the following signs:
-
2.
Cellulitis deemed treatable with orally administered outpatient antibiotics in which either combination of antibiotic is likely to produce a clinical response (Eron Class 1–2)
-
3.
Written informed consent obtained
-
4.
Sixteen years of age or older
-
5.
Fluency in written and spoken English
-
6.
Willing to return for study follow-up or to have the research nurse visit them for follow-up
-
7.
Willing to receive a telephone call from a study investigator
Exclusion criteria
-
1.
Penicillin allergy (self-reported or confirmed)
-
2.
Any cellulitis that treating clinicians deem treatable with intravenously administered antibiotics
-
3.
Any cellulitis of the perineal region
-
4.
Patients who have received more than 24 h of effective antibiotics for the current episode of acute cellulitis
-
5.
Any medical condition, based on clinical judgment, that may interfere with interpretation of the primary outcome measures (e.g. chronic skin condition at the cellulitis lesion site)
-
6.
Immunodeficiency from primary or secondary causes (e.g. corticosteroids, chemotherapeutic agents)
-
7.
Previous history of renal dysfunction or known chronic kidney disease under the care of a nephrologist
-
8.
Previous history of liver dysfunction (defined as chronically deranged liver function tests elicited from medical notes or history)
-
9.
Suspected or confirmed septic arthritis
-
10.
Suspected or confirmed osteomyelitis
-
11.
Infection involving prosthetic material
-
12.
Pregnant or lactating women
-
13.
Patients with a previous history of flucloxacillin-associated jaundice/hepatic dysfunction
-
14.
Patients with a previous history of methicillin-resistant Staphylococcus aureus (MRSA) colonisation/infection
-
15.
Patients with lactose intolerance diagnosed by a medical professional
Interventions
Patients will be randomised in a blinded fashion to either orally administered flucloxacillin 500 mg and placebo, or orally administered flucloxacillin 500 mg and phenoxymethylpenicillin 500 mg, both four times daily for 7 days.
Determination of sample size
The sample size per trial arm was calculated based on an assumed treatment success rate of 85% with orally administered flucloxacillin and phenoxymethylpenicillin, a non-inferiority threshold Δ = 12.5% and α = 0.025 (as this is a non-inferiority study). Sample sizes were calculated using SAS v9.3 (SAS Institute Inc., Cary, NC, USA). Given the preferred study power of 90% for non-inferiority trials and a clinical evaluability rate of 80%, it is estimated that a minimum sample size of 207 in each treatment arm is required (n = 414).
The non-inferiority margin was based on a combination of statistical reasoning and clinical judgement. The non-inferiority margin of 12.5% was considered clinically acceptable based on best current evidence [3]. Additionally, the Infectious Diseases Society of America has recommended a non-inferiority margin of between 10 and 15% for RCTs of antibiotic treatment for cellulitis [4]. Furthermore, current RCT evidence indicates that the efficacy of the comparator intervention (i.e. dual therapy with flucloxacillin and phenoxymethylpenicillin) is 83–87% efficacious. Our sample size for each trial arm was calculated based on an assumed treatment success rate of 85%, which is the midpoint of this 83–87% success range.
Randomisation and blinding
Patients will be stratified by site and type of infection and allocation will be blocked using random permuted blocks of varying size of 2, 4 and 6. A person independent of the trial will create a randomisation plan in advance of the treatment packets being made up. This plan will contain a list of packet numbers and whether the packet number corresponds to study arm A or B. The treatment packet of study medications will be made up by a pharmacist before commencement of the study. Following this, all packets will be assigned a label by the pharmacist corresponding to the packet numbers above. The pharmacist, in addition to the outcome assessors and patients, will be blinded as to which patient receives which treatment pack.
A password-controlled Excel file will be setup (by a person independent of the trial). A research nurse will enter the patient identification number in the Excel file and the packet number to be given to the patient will be revealed. The user of the Excel file is blinded as to which arm of the study the patient is assigned to. The password-controlled Excel file will have built-in validation ensuring that patient identification numbers are entered in strict sequential order and cannot be duplicated or skipped in error.
Individual, sealed patient-number envelopes will be kept in a secure locked press for emergency unblinding access. Contact numbers of the chief investigator (AW) and another study investigator (MQ) will also be available in the circumstances when unblinding is required.
The DSMC may be unblinded to individual study treatment assignments, as needed, to adequately assess safety issues.
Outcome measures/endpoints
Primary outcome measure
The primary outcome measure is investigator-determined clinical response measured at the TOC visit. Measurement of this outcome will be as follows:
-
1.
A suitably trained member of the study team will determine clinical cure at the TOC visit. This is a clinically determined response to treatment based on the judgement of a suitably trained member of the study team
-
2.
Definition of clinical cure:
-
(a)
No treatment failure at any previous visit
-
(b)
Resolution or minimal presence of the following signs and symptoms from the baseline assessment based on the study investigator’s clinical assessment:
-
1.
Erythema
-
2.
Swelling
-
3.
Tenderness
-
4.
Induration
Follow-up assessments will be in person. In the unlikely event that this is not possible, the research nurse will contact the patient by telephone and obtain as much follow-up information from the patient as possible.
Secondary outcome measures
The principal secondary outcome measure is a ≥ 20% reduction in lesion surface area on day 2–3 after enrolment compared to the baseline visit. In addition, the following secondary outcome measures will be assessed:
-
Clinical treatment failure at each follow-up visit
-
Adherence and persistence of trial patients with outpatient antibiotic therapy at EOT (day 8–10 post enrolment)
-
HRQoL assessments at each follow-up visit
-
A pharmacoeconomic assessment of cost per QALY
1. ECR measured on day 2–3
ECR measured on day 2–3 is defined as ≥ 20% reduction in the lesion surface area from that which was measured at enrolment. Objective measurement of lesion size is the US FDA-recommended outcome measure of choice. In order to enhance the validity of the study findings, an objective measure of the percentage decrease in the diameter of infection at day 2–3 will be utilized as a secondary outcome measure. Surface area measurement will be achieved by multiplying the vertical and horizontal diameter of the area of lesion.
2. Clinical treatment failure
Any patient outcome designated as a clinical treatment failure at any time before and including the TOC visit will be categorised as a treatment failure.
3. Adherence and persistence of trial patients with outpatient antibiotic therapy
Medication adherence will be measured at EOT in various ways:
-
1.
By self-report
-
(a)
Have you taken all of your antibiotic doses as prescribed? (yes/no)
-
(b)
If not, how many times would you estimate you did not take your doses as prescribed?
-
(c)
Did you finish your antibiotic course?
-
(d)
If not, when did you stop? (day 0–7)
-
2.
By counting the number of unused study medications at the EOT visit. Patients will be asked to bring their medication with them to the EOT visit so that a pill count can also be performed
-
3.
By using an electronic monitoring system (MEMS® cap). The MEMS® cap will be used to measure adherence in a subset of 100 patients assigned to either treatment arm at the Beaumont Hospital study site only. The MEMS® cap will be fitted to the dispensed medication bottle. Patients will be informed that the cap will monitor each opening of the bottle and, as such, they should only open the bottle when they are about to ingest a dose. Any other openings should be recorded and the research team should be informed. The MEMS® cap should be returned with the clinical trials supplies on the follow-up visits
Medication persistence may be defined as the duration of time from initiation to discontinuation of therapy [5]. Persistence will be measured using the MEMS® cap in the subgroup of 100 patients. It will be measured by determining from the MEMS® cap report the time (in days) from the start day of treatment to the day of the last dose taken without regard for any missing doses or days in between. Missing doses and days will be described separately in descriptions of adherence.
4. HRQoL for economic evaluation
The Foundation for the National Institutes of Health Biomarkers Consortium Project Team and the FDA have explicitly indicated that patient-reported outcome measurements, or how a patient “feels and functions,” are important, required by regulation, and should be measured at early and late time points. However, no RCT included in the Cochrane review of cellulitis treatment assessed HRQoL [6]. The planned study aims to measure HRQoL using the EQ-5D-5L instrument and also to validate a novel disease-specific HRQoL instrument, the ESTI score, in order to address these knowledge gaps. Validation of the ESTI score may also inform future health care decisions about more costly interventions for the treatment of cellulitis. In the meantime, a large proportion of the economic cost of skin and soft tissue infections (SSTIs) caused by community-acquired methicillin-resistant Staphylococcus aureus (CA-MRSA) stems from losses in patient productivity [7]. However, the economic impact of treating ABSSSI for patients presenting to EDs is unknown; by performing a pharmacoeconomic analysis we will provide useful insights regarding the economic impact of treating this relatively common ED condition.
The EQ-5D-5L has been selected as the most appropriate measure of HRQoL. It is a standard measure of self-reported HRQoL which includes five domains: mobility, self-care, usual activities, pain/discomfort and anxiety/depression. The ESTI score will also be measured and mapped against the ED-5D-5 L as a robustness check and this outcome will be used as an alternative way to measure QALYs. The EQ-5D-5L, the 12-Item Short Form Health Survey (SF-12) and ESTI score will be used to obtain patient reports of HRQoL and then used in the estimation of QALYs. The (SF-12) instrument is routinely used as a sensitivity check for quality of life within trials.
5. Pharmacoeconomic analysis of cost per QALY
Although the cost of treating cellulitis may be relatively low, it is important to understand the full health outcomes associated with acute cellulitis in order to establish complete cost-effectiveness for resource allocation decisions. We will perform a within-trial economic evaluation of the cost per QALY gained from the use of orally administered flucloxacillin alone compared with using combination therapy of orally administered flucloxacillin and phenoxymethylpenicillin over a 1-month time horizon from the perspective of the health care payer (direct costs). The QALY measure captures the impact of a treatment on a patient’s length of life and also the impact on their HRQoL and is widely used in health economics as a summary measure of health outcome, which can inform health care resource allocation decisions.
In a secondary analysis the perspective will be extended to consider costs related to the intervention falling on the patient and the government. Costs will be estimated from recorded resources used during the trial, and from health care utilisation and absenteeism/productivity (for the broader perspective) data collected from patients using questionnaires administered at all time points. The EQ-5D-5L, (SF-12) and ESTI score will be used to obtain patient reports of HRQoL and used in the estimation of QALYs. The health economics study aims to identify the within-study incremental cost-effectiveness ratio (ICER; i.e. the costs and benefits of the treatment compared to the control group). Cost per QALY will be the main outcome derived.
Data collection
An electronic Case Report Form (eCRF) will be used to collect all data. For a paper version of this CRF (paper CRF or pCRF; see Additional file 1). Briefly, in addition to the outcome measures specified above, demographics, including age, gender (male/female), ethnicity (White/Black/Asian/Other) and current smoker (yes/no), vital signs including blood pressure, heart rate, temperature, level of alertness, and previous medical history (including history of cardiac disease, diabetes, vascular disease, venous insufficiency, any other diseases) and Charlson Comorbidity Index (CCI) will be collected.
Data on health resource use costs and quality of life will also be collected (see Additional files 2 and 3). The latter will be used to calculate QALYs for the pharmacoeconomic analysis. The CRFs will include a health resource use questionnaire that will be adapted from the standard Client Service Receipt Inventory (CSRI) that is used in trials to collect information about extra resource utilization (e.g. general practitioner visits), medication use, work status and payment of health costs. Patients will receive this at each time point in the trial. Unit costs will be retrieved from Irish and United Kingdom (UK) reference costs. Medication costs will be taken from the Health Service Executive (HSE) Primary Care Reimbursement Service database. The HSE is the statutory provider responsible for all the public health and social services in hospitals and in the community in the Republic of Ireland. Life years gained will be extracted from the published literature. Costs will also include treatment costs. The costs to patients, in terms of time input and travel expenses over the course of the trial, will be estimated. These costs will reflect real-life expenditure post trial. There are still no reference costs available in Ireland (similar to the National Health Service unit costs and Personal Social Services Research Unit costs in the UK), so we will collect unit cost data for relevant resource items from local data sources and expert opinion, or from similar utilisation in the UK if none are available in Ireland. All costs will be up-rated to 2017 prices using the health component of the consumer price index in Ireland. Pounds will be converted to euros using the 2017 exchange rate from the Central Bank of Ireland.
All data being entered into the eCRF will be validated according to agreed criteria appropriate to the data type (i.e. upper and lower limits on numeric values and dates, drop-down lists of valid responses or radio-button selection for multiple option items, dependency-driven data items will only be collected where necessary and relevant).
Analyses sets/populations/subgroups
For the analysis of the primary outcome both intention-to-treat (ITT) and per-protocol (PP) analyses will be performed. Initially, ITT analyses will be performed and will include all patients randomised to the trial regardless of whether they have taken the study drug or not. PP analysis will be defined as in a previous randomised controlled trial for uncomplicated skin abscesses conducted by Talan et al. in the United States [8]. It will include participants who either took ≥ 75% of the total doses of study drug or placebo during the first 5 days and had an in-person test-of-cure visit or were determined to have had clinical failure before the test-of-cure visit and received ≥ 75% of the doses provided during the first 48 h of the treatment period. In a non-inferiority trial setting it is suggested that a PP analysis may be more appropriate than ITT since it is more likely to reflect actual differences between the two treatments [9]. Furthermore, ITT analysis may be interpreted as being too liberal in a non-inferiority trial and may bias toward making the two treatments appear similar. As a result, both an ITT and PP analysis will be performed on the resulting data to assess non-inferiority. Non-inferiority will only be declared if both ITT and PP analysis support non-inferiority.
Handling of missing data values
Any missing data or data anomalies will be communicated to the study site(s) for prompt clarification and resolution. It is not anticipated that there will be a huge amount of missing data. However, in the unlikely event that there is more than 10% of data values missing, missing values will be imputed, if possible, using a suitable imputation method.