Study setting and participants

All adult patients with CAP are screened and enrolled at emergency departments or medical wards in seven tertiary care hospitals of Switzerland within 24 hours; on weekends within 36 hours of hospitalization. Eligible patients are asked for their informed consent and are then centrally randomized to either 50 mg prednisone or to placebo, taken orally for 7 days in a ratio of 1:1. If the patient is not able to give informed consent at the time of hospitalization due to severity of CAP, informed consent from next of kin and an independent physician is obtained. In these cases, informed consent from the patient himself has to be obtained as soon as possible. Patient recruitment started in December 2009 and is planned to be completed in April 2014.

Inclusion criteria are age of 18 years or older and hospitalization with CAP defined by a new infiltrate on chest radiograph and the presence of one or several of the following acute respiratory signs or symptoms: cough, sputum production, dyspnea, core body temperature≥38.0°C, auscultatory findings of abnormal breath sounds and rales and leukocyte count>10 or<4 G/L [21].

Exclusion criteria are permanent inability for informed consent, active intravenous drug abuse, acute burn injury, gastrointestinal bleeding within the past three months, known adrenal insufficiency, a condition requiring more than 0.5 mg/kg/day prednisone equivalent, pregnancy or breast feeding, and severe immunosuppression defined as previously known infection with HIV and a CD4 cell count below 350 G/L, immunosuppressive therapy after solid organ transplantation, neutropenia<500 G/L or a neutrophil count of 500 to 1,000 G/L with an expected decrease to values<500 G/L in patients under ongoing chemotherapy, cystic fibrosis and active tuberculosis.

Randomization and blinding

Allocation of patients to either intervention in this randomized placebo-controlled trial is concealed due to a pre-specified computer-generated randomization list which is kept centrally at the pharmacy of the main study center. Randomization between the two arms are 1:1 with variable block sizes of 4 to 6 and are stratified at the time of study entry by study center.

Generator and executor of randomization are separated. Patients are randomly assigned to receive a prepared set of study medication that contains 7 tablets of 50 mg prednisone or placebo. The drugs are prepared prior to the initiation of the study and packed by the Pharmacology Department, University Hospital, Basel, according to the randomization list. The numbered study medications are then delivered to the study team. By this method, study team including data analysts, patients and treating physicians will be blinded to group allocation. Unblinding will be performed only upon request of a treating physician or of the data safety and monitoring board (DSMB).

Study protocol

The study flow chart is shown in Figure 1. After informed consent is obtained, baseline blood is drawn and a low-dose adrenocorticotropic hormone (ACTH) test is performed (measurement of basal cortisol followed by cortisol after 0.001 mg of tetracosactid (Synacthen®)). Study medication is started after the ACTH test, and timing in relation to start of antibiotics will be monitored. Patients are evaluated for the primary endpoint of clinical stability every 12 hours during the hospitalization. All patients are being treated according to current CAP guidelines [22,23]. Antibiotic therapy guided by an established procalcitonin algorithm [24] is encouraged. The time of discharge is not determined by the study investigators, but based on the decision of the treating physician team.

Open in a separate window

Figure 1

Study flow chart.

Laboratory assessment

The blood samples drawn on days 1, 3, 5, 7 and at discharge are immediately centrifuged and stored at -70°C until further processing. All laboratories of the participating centers are internationally validated and standardized. Cortisol is measured immediately. Routine laboratory tests in both groups include procalcitonin, C-reactive protein (CRP), white blood cell count, sputum and blood cultures, nasal swab for virus multiplex PCR, Legionella and pneumococcal antigen in urine. Other laboratory markers as well as genetic markers will be measured in batch analysis from stored serum or plasma samples, respectively, after the study is terminated.

Data collection

Baseline data are collected by electronic case report forms (eCRF) compliant to good clinical practice (GCP) and contain date and time of randomization, birth date, sex, medical history items (dyspnea at rest or after having taken a flight of stairs, cough, smoking history (pack-years) and status (pack per day), relevant co-morbidities, degree of autonomy at home, clinical items of pneumonia, pneumonia severity index (PSI), extent of pneumonia, underlying chronic obstructive pulmonary disease (COPD) as assessed by patient history and medical documents.

In both groups, hospitalized patients are reassessed clinically every 12 hours (including vital signs) and blood sampling on days 3, 5 and 7 including 4 measurements of glucose levels on days 3, 5 and 7, respectively. Details on dosing of all prescribed antimicrobials during the study period are recorded. In hospitalized patients, hospital discharge is recommended if oral intake is feasible, vital signs are stable≥24 hours, and no evidence of acute serious co-morbidity that necessitates hospitalization is found. A final blood collection is foreseen on the day of discharge.

On days 1 and 5, the patient has to complete a short questionnaire on respiratory symptoms.

Any additional diagnostic testing in patients in both groups is at the full discretion of the treating physician.

Follow-up

Structured follow-up telephone interviews for secondary outcomes after discharge are performed on days 30 and 180. In case the patient cannot be contacted or cannot give sufficient information, the primary care physician is interviewed.

Follow-up items include assessment of adverse events, such as infections, recurrent pneumonia, re-hospitalization, new diabetes or insulin dependence, new hypertension, and the standardized questionnaire on respiratory symptoms that was used on days 1 and 5.

Endpoints

Subgroup analyses

We plan to explore treatment effects according to the following parameters:

• Above versus below the median initial CRP and maximal CRP (hypothesis: patients with higher CRP benefit more from adjunct corticosteroid therapy)

• Age<70 years versus>70 years (hypothesis: patients above 70 years benefit more, as older-aged patients present with more severe CAP and have a higher mortality [27]. So far, there is no age-related data on adjunct glucocorticoid therapy, but in severe CAP there is more evidence for benefit of glucocorticoid therapy [16])

• Severity of CAP (low-risk group: PSI I to III versus high-risk group: PSI IV to V (hypothesis: patients with more severe CAP and supposedly more severe inflammation benefit more)

• Microbiological subgroups with versus without bacteremia (hypothesis: patients with bacteremia benefit more)

• With versus without underlying COPD according to the Global initiative for chronic Obstructive Lung Disease (GOLD) classification (hypothesis: patients with COPD benefit more)

• COPD patients with GOLD category I to II versus III to IV (hypothesis: patients with GOLD category III to IV benefit more)

Sample size considerations

This study is designed to show superiority of the steroid group compared to placebo treatment. Our primary hypothesis is that adjunct corticosteroid treatment in patients with CAP will result in a reduced time to clinical stability compared to placebo treatment. To estimate the frequency of our primary endpoint, we used data of CAP patients from our completed studies [24,28,29] which were performed in the same study centers. Based on these previous trials, we assumed a mortality rate of 10% in the placebo group and 7.5% in the corticosteroid group over 14 days of follow-up with a proportion of 75% survivors being clinically stable after 7 days in the corticosteroid group. Estimating a decrease in the risk of non-stability after 1 week among survivors by 25% through adjunct corticosteroids, a sample size of 800 patients followed for at least 14 days after randomization will be needed to achieve a statistical power of 85% and a sample size of 700 patients to achieve a statistical power of 80% (2-sided type 1 error of 5% in both scenarios). We intend to recruit 800 study patients from December 2009 to April 2014.

Intention-to-treat and per-protocol population

Following the intention-to-treat principle, all patients receiving at least one dose of study medication will be included in the analysis with group allocation as randomized; patients violating inclusion criteria or meeting exclusion criteria due to information that was not available at study entry will be excluded post randomization in a blinded manner [30]. Criteria for exclusion of patients from the per-protocol analysis are, in addition, withdrawal of informed consent or non-compliance with study medication or the concomitant procedures. Therefore, patients stopping or pausing the study medication at any time point, be it due to corticosteroid contraindications or due to active indication for glucocorticoid treatment, will not be considered in the per-protocol analysis. A patient will not be followed up if he withdraws informed consent for all study measures; in all other cases, follow-up will be performed as planned.

Statistical analysis

We will enter all relevant clinical and laboratory data obtained by interview, clinical tests, and reviewing of the medical records into an online database. Statistical Analysis System (SAS® Institute, Cary, NC, USA) and R for Windows (R Foundation for Statistical Computing, http://www.r-project.org) and STATA 9.2 (Stata Corp, College Station, TX, USA) will be used for data analysis.

The primary analysis population is the full analysis set which includes the intention-to-treat population. Every effort will be made to minimize the number of losses to follow-up by recording phone numbers before discharge, thereafter by repetitive follow-up calls by home or cell phone contact, or alternatively by contacting the primary care physician. For the primary outcome of time to clinical stability, we consider losses to follow-up in this sample of hospitalized patients as unlikely. If losses to follow-up do occur, patients will be censored at time of last assessment for clinical stability. A secondary analysis population, the per-protocol population, will exclude patients not treated according to the protocol (that is protocol violators). For the primary endpoint, we will calculate an unadjusted hazard ratio and 95% confidence interval by means of a Cox proportional hazards model. The primary objective is to show superiority of the corticosteroid group at a 2-sided 5% alpha-level with 85% power following the intention-to-treat principle. As a sensitivity analysis, the primary analysis will be repeated on the per-protocol patient population. As a further sensitivity analysis, a multivariable Cox proportional hazards model will be fitted with treatment group, patient age, and PSI scores as independent variables. This model will adjust the treatment effect for the potentially important confounders age and PSI score. Moreover, we plan to investigate whether the treatment effect varies among different degrees of severity of CAP (low-risk group: PSI I to III, and high-risk group: PSI IV to V), between patients with and without underlying COPD, and between bacteremic versus non-bacteremic patients. COPD will be defined by post-bronchodilator spirometric criteria, according to the GOLD guidelines [31], as a FEV1/FVC ratio below 70% and the severity categorized according to the GOLD criteria. In patients with a clinical history of COPD or smoking, lung function at the time of inclusion will not be mandatory. We will conduct the pre-specified subgroup analyses by including appropriate interaction terms in the above described multivariable Cox proportional hazards model.Complete case analyses will be used for secondary endpoints. For all secondary endpoints, we will calculate unadjusted and adjusted estimates of the effect size and corresponding 95% confidence intervals using linear, logistic, or Cox proportional hazards regression (as appropriate). Figure 2 shows the CONSORT diagram of the trial.

Open in a separate window

Figure 2

CONSORT diagram of the trial.

Study organization

All adverse events of category III or more according to Common Terminology Criteria for Adverse Events v4.0 [32] are recorded. Adverse events will be monitored for every subject participating in the study and attributed to the study medication and reported to the respective authorities according to law if necessary.

The trial is supervised by an independent and blinded DSMB that is not involved in the design and conduct of the trial, and has no affiliation with the sponsor. The DSMB is comprised of experts from different disciplines and supervises the recruitment of patients, patient flow and patient interviews. The board has access to the randomization code in case of need. In case of serious problems or adverse events that may occur during the trial, the board will review the preplanned interim analysis and inform the principle investigators whether the trial may be continued or has to be stopped early or it may require modifications of the protocol.

A blinded safety interim analysis was performed after 50% of patients had completed the 180 day follow-up. The DSMB recommended continuation of the study without changes to the protocol.

In accordance with applicable regulations and GCP, a study monitor is periodically controlling study procedures.