No clearcut guidelines exist for when cellulitis requires treatment with intravenous antibiotic other than in case oral antibiotics fail. In this study the decision whether intravenous antibiotics were required was left to the attending doctors in the emergency department who assessed the patient.
No validated objective measures seem to exist of when cellulitis is improving or when patients can be switched from intravenous to oral antibiotics. We chose as our primary outcome measure the time to when the cellulitis failed to advance. This outcome has been used in one previous study.8
Other outcomes recorded included the total numbers of days when patients received intravenous antibiotics and oral antibiotics, and calendar days in hospital or looked after by the home care team. The decision when to switch patients from intravenous to oral antibiotics was left entirely to the attending doctor in the hospital or home. We recorded patients' transfers from home to hospital and the reasons for transfer. We kept a record of all serious complications experienced by patients. We used questionnaires to assess patients' level of functioning and pain as well as satisfaction with their care.
Christchurch Hospital serves the whole metropolitan area of Christchurch (population in 2001 was 318 000), and all acutely referred patients are treated there. We informed all general practitioners in Christchurch and emergency department staff at Christchurch Hospital of this trial before it started.
We recruited participants from patients with cellulitis who were attending Christchurch Hospital's emergency department, whether self referred or referred by their general practitioner or a general practitioner after hours. Patients who were considered to require intravenous antibiotic treatment for cellulitis by the emergency doctor and who met the eligibility criteria received an invitation to take part in the trial.
Patients were eligible for the trial if they had clinical signs of cellulitis, were assessed as requiring intravenous antibiotic treatment because of severity of cellulitis or failure of oral antibiotic treatment, were 16 years or older and mentally competent to give informed consent, had a telephone at home and a caregiver nearby, and were currently resident in the Christchurch metropolitan area.
Exclusion criteria were pregnancy; treatment with intravenous antibiotics for cellulitis of the same site in the preceding month; two or more signs of systemic sepsis (temperature > 38°C or < 36°C, heart rate > 90/min, respiratory rate > 20/min); and a blood count showing a white cell count above 12×109
/l or less than 4×109
/l and more than 0.1×109
/l immature neutrophils.9
Other possible exclusion criteria were signs of severe cellulitis or serious comorbidities such as cellulitis of the face, hands, or over joints; presence of tissue necrosis, severe lymphangitis, blistering, or a very large affected area; comorbidities such as immunosupression, peripheral vascular disease, obesity, alcoholism, or severe diabetes. The more of these relative exclusion criteria were present the more hospital admission was recommended.
Routine blood tests were not required, and the criteria for exclusion were deliberately kept flexible as ultimately the staff in the emergency department often had to make a subjective judgment about the suitability of a patient for entry into the trial. This decision was made independently from the investigators conducting the trial, and junior staff in the emergency department always conferred with consultant staff in making the decision to enrol patients in the trial. Between the hours of 8.00 and 22.00, a member of the study team visited the patient in the emergency department and, after the patient had read the trial information and consent sheets, obtained informed consent. Outside this time patients received an initial dose of intravenous cephazolin and were looked after in the emergency department's observation ward until the following morning when study staff obtained informed consent.
Once a patient had given consent he or she was assigned a unique study number, and allocation to home or hospital treatment was determined by phoning an off-site coordinator who kept the randomisation list and assigned each study number to either home or hospital treatment. The randomisation list was produced by SAS code from the SAS statistical package (SAS Institute, Cary, NC 27513-2414, USA) using randomly allocated block sizes with a maximum of 20. In each block, equal allocations were made to the two arms of the trial.
The study team collected information on the participants in the emergency department, including demographic information (sex, date of birth, ethnicity, address, occupation, community card status); details of any current or recent use of antibiotics, the location of cellulitis; and the presence of any skin necrosis, lymphangitis, blistering, or ulceration.
The researcher drew an indelible line with a marker pen around the peripheral margin of the cellulitis and dated this for comparison on following days.
Before leaving the emergency department, every participant received his or her first intravenous dose of 2 g of cephazolin. If renal impairment was suspected or known, the creatinine concentration was measured and the dose adjusted. Those participants randomly allocated to hospital treatment were then admitted to a hospital ward under the care of the on-call medical team who managed the subsequent clinical treatment, including the choice of ongoing intravenous antibiotic. Participants allocated to hospital treatment were visited each day by the study team to record clinical progress.
Patients who were randomly allocated to community treatment continued with 2 g of intravenous cephazolin (modified in renal impairment) twice daily. Their own general practitioner or a general practitioner from the community care team visited them daily for medical review, and community care nursing staff attended twice daily to monitor the cellulitis and administer intravenous antibiotics. Research staff reviewed community and hospital participant clinical records in all cases. This review included duration of stay, details of antibiotic treatment, and complications.
At entry into the trial and at days 3 and 6, we administered a questionnaire modified from the short form 36 (SF-36) instrument, which focused on functional and physical aspects of health.10
At trial entry we asked patients to respond about their health before the infection, whereas at days 3 and 6, we asked them to respond about their health in the previous 24 hours. We administered questionnaires face to face at trial entry and when participants remained in hospital or by telephone if the patients had left hospital. Patients completed a patient satisfaction questionnaire four weeks after entry into the trial.
The study was designed to have 200 participants, 100 in each arm. With power of 80% and α2 = 0.05, a moderate difference of 0.40 standard deviations was detectable between the two arms for the primary outcome of no advancement of cellulitis. The clinician researchers thought that a difference of up to two days would be acceptable. The standard deviation in each group was not known, but as long as it was less than five days the study had adequate power. We used survival analysis for the main clinical outcomes and, to compare the groups, χ2 tests for contingency tables and t tests for continuous variables. We carried out our analyses in SAS, version 8.02 (SAS Institute, Cary, NC 27513-2414, USA).