Comparing short to standard duration of antibiotic therapy for patients hospitalized with cellulitis (DANCE): study protocol for a randomized controlled trial

The DANCE trial is an investigator-initiated, prospective, multicenter, randomized, investigator-, caregiver- and patient blinded non-inferiority trial, with two parallel groups with an allocation ratio of 1:1. Subjects will initially be enrolled at eight Dutch hospitals: i) Academic Medical Center, Amsterdam; ii) VU University Medical Center, Amsterdam; iii) University Medical Center, Utrecht; iv) Diakonessenhuis, Utrecht; v) Flevoziekenhuis, Almere; vi) Onze Lieve Vrouwe Gasthuis, Amsterdam; vii) Sint Lucas Andreas Hospital, Amsterdam; viii) Slotervaart Hospital, Amsterdam. These sites serve a large, densely populated urban area of the Netherlands. In the case of slow enrolment, additional hospitals will be approached to participate in the study. This study will be conducted in accordance with the Declaration of Helsinki and was approved by the medical ethics committee of the Academic Medical Center, Amsterdam. The Boards of Directors of all participating sites subsequently approved the execution of the study in their centers.

Adult patients meeting the in- but not exclusion criteria are eligible (see below). Cellulitis is defined as warmth, erythema, and induration of the skin and/or subcutaneous tissue, with or without pain (including erysipelas) [33]. Eligible patients are approached by local investigators who will provide them with the patient information folder. Two investigators (DRC and a research nurse) will be responsible for the informed consent process and data collection in every site. They will perform all study procedures. If this proves to be excessively time consuming, additional investigators will be recruited and trained. After providing consent, subjects can leave the study at any time for any reason if they wish to do so, without any consequences. The investigator can decide to withdraw a subject from the study for urgent medical reasons. Once randomized, withdrawn subjects will not be replaced, in accordance with the intention-to-treat principle.

The inclusion criteria are:

The exclusion criteria are:

All participants will receive regular antibiotics (flucloxacillin, or another empiric beta-lactam with activity against S. aureus that is shortly after replaced by flucloxacillin) for the first 6 days of treatment. Treatment will be started intravenously, but a switch to oral medication can be performed at the caregiver’s discretion when the patient has become afebrile and the cellulitis shows improvement. Participants will be assessed by one of the investigators at multiple time points (Table 1).

Participants responding to initial therapy, defined as absence of fever (temperature > 38.0°C) and improvement in cellulitis severity score by day 5–6 compared to the first 24 hours of admission, and able to take oral medication will be randomized into one of two groups. One group will receive 6 more days of flucloxacillin (500 mg q.d.s.), while the other group will receive 6 days of placebo (500 mg q.d.s.). Of note, patients with a creatinine clearance of <10 ml/min or those on haemodialysis will receive 500 mg of trial medication t.d.s. Patients are not randomized if: i) they require ICU admission, concomitant antibiotics, or surgical intervention, or develop an allergic reaction (see exclusion criteria); ii) their blood cultures become positive (excluding contamination); or iii) they do not improve. Other aspects of treatment are according to standard of care. Compression therapy should be encouraged, but is not mandatory, and is at the discretion of the caregiver. Patients will be informed about the correct use of the trial medication. To register adherence, patients will be asked to bring left-over trial medication to their day 14 visit, at which the capsules will be counted and afterwards destroyed.

Standard demographic and other patient characteristics will be collected, including age, gender, non-trial medication use, comorbidity, ethnic background, recreational drug use, smoking, body weight, residential status, and risk factors for cellulitis [6] (i.e. venous insufficiency, lymphoedema, peripheral arterial disease, immunosuppression, diabetes, trauma, tattoos, ulcers, eczema, tinea pedis and burns). The extent of cellulitis (length and width, lesion surface area) and presence of lymphadenopathy will be recorded on pre-determined time points (Table 1). A cellulitis severity score will be used, which measures erythema, warmth, tenderness, edema, ulceration, drainage and fluctuation. Each parameter is scored with a numerical value of 0–3 (none, mild, moderate, and severe, respectively). This has been used previously in a trial on duration of therapy for uncomplicated cellulitis [23].

The two investigators responsible for data collection will standardize these assessments by performing simultaneous assessment at the beginning of the study. When five patients have been scored independently with identical scores, the investigators will start assessing patients on their own. After 100 patients have been assessed, the investigators will again assess five patients together. VAS scores and questionnaires will be collected through a web survey tool. Microbiological culture results and hematology and chemistry lab results will be collected from regular care, when available. Additionally, blood samples will be taken at multiple time points, and skin biopsies and swabs will be taken from selected patients on admission.

The primary outcome is resolution of cellulitis at day 14, defined as disappearance of warmth and tenderness at the site of infection, with substantial improvement in erythema and edema, and without recurrence by day 28, defined as the need for additional antibiotic therapy for cellulitis.

Secondary outcomes are recurrence of cellulitis by day 90; speed of recovery, determined by improvement in cellulitis severity score and through self-assessment of subjective pain and swelling on visual analog scales (VAS) from 0–10; mean health-related quality of life, using the SF-36 and EQ-5D questionnaires; and costs, associated with total antibiotic use and other health-care resource utilisation, measured with modified versions of iMTA’s Productivity Cost Questionnaire (iPCQ) and Medical Consumption Questionnaire (iMCQ).

Blood samples and skin biopsies will be used for biomarker discovery, fundamental research, and skin microbiota analyses.

Patients eligible for randomization will be allocated to either (1) flucloxacillin or (2) placebo in a 1:1 randomization ratio, using the online ALEA® software developed by the NKI-AVL (Amsterdam, NL). This independent central randomization service will create a computer generated random schedule of blocks with random block sizes (maximum block size of 6), stratified for presence of diabetes mellitus and study site. Upon randomization, the software will generate a randomization number for the investigators, and will send an automated e-mail to the appropriate pharmacy with the randomization number and randomization result. Pharmacies will dispense randomly numbered trial medication containers to the patients, depending on the randomization result, working from a confidential list which describes the content (flucloxacillin or placebo) of each numbered container. Flucloxacillin and placebo are in similar capsules, and are administered in the same dosage and duration, leading to blinding of patients, investigators and caregivers. The randomization code will only be broken in case the Data Safety Monitoring Board (DSMB) advises to do so.

When the sample size in each group is 198 (396 total), a two-group large-sample normal approximation test with a one-sided 0.050 significance level will have 80% power to reject the null-hypothesis that 6 days of antibiotics is not inferior to 12 days of antibiotics. This assumes a conservatively estimated treatment failure rate of 20%, and considers a 10% absolute difference in treatment failure rate between the two groups as equivalence limit.

Statistical analysis will be performed using the modified intention-to-treat principle (mITT), including all randomized patients who received at least one dose of the study medication. Missing data will be handled with multiple imputation. Baseline assessments and outcome parameters will be summarized. Continuous variables will be summarized with standard descriptive statistics, categorical variables with frequencies and percentages.

For the primary outcome, the proportion of successful treatments will be calculated for the two groups, and the absolute risk difference (with one sided 95% confidence limit) will be estimated to evaluate the non-inferiority hypothesis (Table 2). In a secondary analysis, the difference in proportion in the two groups will be adjusted for relevant baseline covariates, using a logistic regression analysis. The difference in proportion of successful treatments will be estimated at mean values for the covariates. Differences in secondary outcome parameters will be reported as relative risk or mean difference (with 95% confidence interval), and p-values for statistical significance estimated with the Chi-square test, t-test or Mann–Whitney test, where appropriate. A Kaplan-Meier analysis will be performed to compare the time to relapse (defined as requiring additional antibiotics for cellulitis) between the two groups.

A per protocol analysis will also be performed, for this analysis the following randomized patients of whom the compliance is clear will be included: patients with treatment failure who have received at least 24 hours of study medication, and patients with treatment success who have received at least four days of study medication.

For this analysis treatment success is defined as resolution of cellulitis at 14 days – disappearance of warmth and tenderness at the site of infection, with substantial improvement in erythema and edema –without recurrence by day 28, defined as the need of additional antibiotic therapy for cellulitis [23]. Treatment failure is defined as the persistence or progression of signs and symptoms of the acute process at 14 days after randomization, a recurrence before day 28, or the inability to complete the study owing to adverse events. The response is deemed indeterminate when the patient (i) received less than 80% of the study drug for reasons other than treatment failure, (ii) acquired a concomitant infection outside of the skin requiring antibiotic treatment, (iii) was lost to follow-up, or (iv) died unrelated to the primary diagnosis.

Two subgroup analyses are planned, and statistically tested with interaction effects: one for patients with diabetes mellitus, and one based on the cellulitis severity score. Diabetes mellitus is known to influence outcome in patients treated for complicated skin and skin structure infections, through multiple suggested mechanisms [36]. A high cellulitis severity score reflects an extensive local inflammatory response, which might indicate a higher grade of infection that possibly requires more intensive or prolonged antibiotic therapy.

An independent DSMB has been established, consisting of two infectious disease specialists and a methodologist. They will evaluate mortality after 198 patients have completed the trial. Given the expected low mortality rates [37], the power to detect differences in mortality at this stage will be very low. Therefore, the DSMB will qualitatively evaluate all deaths for their possible relation with the given treatment. Only in case of striking differences between the two arms the DSMB will inform investigators and the medical ethics committee. The frequency of other DSMB meetings will depend on the study proceedings and the judgment of the DSMB chair.

Patient recruitment for the trial will start from the beginning of 2014. Data collection is estimated to finish in 2017.

This study is financed by a grant of the Netherlands Organisation for Health Research and Development (ZonMW) (grant nr 836011024).