Introduction
Background and rationale
Community-acquired lower respiratory tract infections (LRTI) are one of the most common motivations for consultations in the emergency department (ED) and stand as the leading cause of inappropriate antibiotic prescription [
1]. LRTIs span a wide range of diseases, from self-limited acute bronchitis and infectious exacerbations of chronic obstructive pulmonary disease (COPD) to life-threatening pneumonia. Viruses cause the majority of LRTIs and are also identified in a quarter of community-acquired pneumonia (CAP), with an even higher prevalence during the SARS-CoV-2 outbreak [
2,
3].
When assessing patients with LRTIs, the challenge for ED physicians is to identify those with bacterial CAP, who are most likely to benefit from antibiotics. The performance of current tools to diagnose CAP in patients with LRTI is limited. Chest X-ray, the current reference standard for pneumonia diagnosis, has poor accuracy [
4,
5]: based on clinical features and chest X-ray, pneumonia is largely overestimated, as a third of patients have a normal CT scan [
6]. The low diagnostic accuracy of existing tools is one of the causes of inappropriate antibiotic prescriptions and excessive utilization of costly resources (blood tests, radiology, and microbiology) among patients with LRTIs [
7‐
11]. Although 40% of patients with LRTIs have CAP in the ED, 60 to 80% of patients and almost all those with CAP receive antibiotics [
3,
12,
13]. Besides their side effects, antibiotic overuse alters the microbiome and selects for antibiotic resistance [
14,
15].
Several diagnostic tests can assist in identifying patients with LRTI which require antibiotics. Lung ultrasound (LUS) can be performed quickly at the bedside without radiation and has a better diagnostic performance than chest X-ray to detect infiltrates. Recent meta-analyses have shown that LUS has an excellent sensitivity (92–94%) and specificity (74–96%) in diagnosing CAP in adults ED patients using chest CT as gold standard [
16‐
18]. However, based on a recent review, 25% of pneumonias are viral and cannot—based on imaging alone—be distinguished from a bacterial pneumonia [
2]. Procalcitonin (PCT) is a host inflammatory biomarker which is usually elevated in bacterial and/or severe infections [
19]. While PCT can be used to safely guide antibiotic use, its impact on their prescription remains controversial mainly due to variable physician adherence to PCT guidance [
13,
20,
21]. If none of these tools in isolation is sufficient to optimize antibiotic prescription, a combined approach could improve diagnostic performance, ensure patients’ safety, and maximize clinicians’ adherence to guidance.
To overcome the limited performance of guideline-recommended diagnostic approach and the shortcomings of previously tested interventions for the management of LRTIs, we provide ED physicians with a novel simple clinical management algorithm: the PLUS algorithm. The PLUS algorithm integrates a clinical predictive score for CAP (Van Vugt score), LUS, PCT, and, in case of discordant results between lung ultrasound and PCT, a clinical severity score (DS-CRB-65). The purpose of this algorithm is to improve the identification of patients with CAP requiring antibiotics and decrease unnecessary prescriptions in adult patients with LRTIs managed in Swiss EDs. It guides the bedside decision-making for antibiotic prescription and helps physicians to safely withhold unnecessary prescriptions in adult patients with LRTIs.
Objectives
The primary safety objective of the Procalcitonin and Lung UltraSonography based antibiotherapy in patients with Lower rESpiratory tract infection in Swiss Emergency Departments (PLUS-IS-LESS) study is to demonstrate a non-inferiority of the PLUS algorithm, in terms of clinical failure by day 28 when compared with usual care, among patients with LRTIs in the ED. The co-primary efficacy objective is to show a 15% reduction in the proportion of patients with LRTIs prescribed an antibiotic by day 28 in the intervention group compared to the usual care group.
Secondary objectives of the study are to compare between the intervention and control groups the quality of life (the inconvenient nature of CAP-related symptoms) on day 7, day 28, and day 90, the duration of ED stay, the rate and duration of hospitalization, the proportion of patients prescribed an antibiotic for a new respiratory infection between day 28 and 90, as well as the proportion of patient with antibiotic-related side effects and Clostridioides difficile infection. In addition, we will evaluate the acceptability and feasibility of the intervention through identification of barriers and facilitators for patients and physicians, compare quality-adjusted life years (QALYs), derived from responses to the EQ-5D-5L questionnaire (euroqol.org), assess the incremental cost-effectiveness of the intervention as compared to usual care, develop advanced automatic LUS image analysis method using machine learning to assist in LUS diagnosis and risk stratification, assess the proportion of physicians reaching proficiency in LUS image/video acquisition and interpretation after the training module, and evaluate the diagnostic performance of physician “gestalt” and Van Vugt prediction score for CAP diagnosis.
Methods: Data collection, management, and analysis
Data collection
Table
6 summarizes the schedule of enrolment, intervention, and assessments.
Table 6
Schedule of enrolment, intervention, and assessments (as per SPIRIT [
22])
At enrolment
Study data will be collected (from the patient and/or the medical file) by the study team in an eCRF developed on the electronic data collection (EDC) system REDCap®, hosted at Lausanne University Hospital [
35,
36]. Data collected at inclusion includes demographic characteristics, comorbidities, symptoms and their duration, vitals and clinical signs, diagnostic tests required by the ED physician for LRTI management of the included patients (chest X-ray, CT-scan, LUS, CRP, PCT, white blood cell count, blood chemistry, blood gas, microbiology tests), destination on ED discharge, and antibiotic prescription. During the intervention period, the data collection will also include PCT results and LUS interpretation. Furthermore, physicians will record all their LUS images and videos following the aforementioned procedure in a secured drive.
A collection of biological samples (blood, respiratory, and urine) is performed parallel to the intervention trial as a study data-linked biobank. This includes a collection of nasopharyngeal and oral swabs, sputa if available, whole blood, plasma, PAXgene™ blood RNA system tubes, and urine samples at inclusion in all study centers. Furthermore, urine samples are collected on day 7 in the University Hospital of Lausanne and the Cantonal Hospital of St Gallen.
During follow-up
Study data will be collected by phone on days 7, 28, and 90. The study team will perform the follow-up phone calls. If the patient is admitted on the day of follow-up, the study team will interview the patient. They will also use the patient’s medical file if relevant. Data collected at follow-up includes mortality, hospitalization, use of oxygen therapy or non-invasive ventilation, presence of local complication of pneumonia, use of antibiotics, health care resources, LRTI-related symptoms, and evaluation of quality of life.
Data management
Study data will be collected via an electronic case report form (eCRF) and managed using REDCap®. Trial and participants’ data will be handled with uttermost discretion and is only accessible to authorized personnel who require the data to fulfil their duties within the scope of the study. On the CRFs and other study-specific documents, participants are only identified by a unique participant number (3 to 4 letters for the site followed by 3 digits for the patient). Coding is generated by using the eCRF software. Biological material in this study is not identified by participant name but by a unique participant number. Biological material is appropriately stored in a restricted area only accessible to the authorized personnel.
Each recording and transcript collected during the qualitative interviews will be identified by the patient’s or physician’s unique participant number. To further ensure confidentiality, transcripts will be coded for any name or place that could allow to identify a participant.
Quality management and monitoring
Monitoring
The monitoring activities are coordinated by the Clinical Research Center (CRC) Lausanne. The Clinical Trial Center Zurich (members of the Swiss Clinical Trial Organization (SCTO) and CTUs Network) is contracted by Lausanne to perform monitoring activities at some local sites.
A risk-adapted monitoring strategy were developed according to the SCTO guidelines (Guidelines for Risk-Based Monitoring, version 3.0). The monitoring strategy (nature and extent of the monitoring activities) were described in a monitoring plan. Site initiation visits were conducted on each site before the start of the trial. Then interim monitoring visits (6 per site) will be performed on site and focusing mainly on safety and patient eligibility. A closeout monitoring visit per study site will be performed at the end of the trial to ensure all pending actions are resolved and study documentation is ready for archival.
Quality controls
The quality controls include the evaluation of study center compliance to the protocol. Non-adhesion to the algorithm will be escalated to the sponsor as a study deviation triggering corrective action from the investigators. Furthermore, non-adherence to the algorithm recommendation will trigger automatic alerts sent to the sponsor for direct feedback to the clinician in charge. There will be a monthly follow-up of deviation from the research proposal. This includes an implementation evaluation of the intervention to detect problems, which may be corrected.
Safety monitoring
In our study, an adverse event (AE) is defined as any untoward medical occurrence in a participant which does not necessarily have a causal relationship with the trial procedure. A serious adverse event (SAE) is defined as any untoward medical occurrence that results in death or is life-threatening, requires in-patient hospitalization or prolongation of existing hospitalization, results in persistent or significant disability or incapacity, or causes a congenital anomaly or birth defect. Occurrence of SAEs from the first participant visit until the last follow-up phone will be actively sought. All SAEs occurring during the study will be documented in the eCRF and followed until resolution. SAEs identified as related to the intervention will be notified to the Ethics Committee according to regulatory requirements, as well as immediate safety and protective measures that should be taken during the conduct of the study.
Protocol amendments
The Sponsor is authorized to amend the protocol. All important protocol modifications will be first discussed within the Steering committee and then communicated to the relevant parties (local investigators, EC, trial registry) by the Sponsor. Substantial amendments will only be implemented after approval by the EC, whereas non-substantial amendments are communicated by the Sponsor to the EC within the annual safety report. Under emergency circumstances, deviations from the protocol to protect the rights, safety, and well-being of patients may proceed without prior approval of the EC. Such deviations shall be documented and reported by the Sponsor to the EC as soon as possible. Amended protocols will be sent to the study sites for filling in the Investigator Site File, and training on new documents will be documented on site.
Analyses
Sample size
The study is designed to demonstrate non-inferiority of the intervention in term of clinical failure occurrence and superiority in term of reducing antibiotic prescription. We plan to include 1530 patients in 9 study centers. Every study center will recruit 170 patients over about 2 years.
According to a stepped-wedge cluster-randomized design with a control period plus nine switch periods (steps) and the same number of clusters (one switch to intervention at each step), a mean number of 15 patients per unit and per time period will guarantee a 80% power (one-sided type I error α = 0.05) to prove non-inferiority regarding clinical failure (safety), if the proportion of clinical failure is 20% in both groups, with a non-inferiority margin of 10%, and assuming an intra-cluster correlation (ICC) of 0.2 and a coefficient of variation (cv) of the number of patients across units of 0.3. With 15 patients per unit and per time period, we would obtain a total sample size of 1350 (15 × 9 × (9+1)) patients. Further considering a 10% of lost to follow-up, we anticipate 15/0.9 = 17 patients per unit and per time period, leading to our final sample size of 1530 (17 × 9 × (9+1)) patients.
In our calculation, the ICC of 0.2 is based on results of a previous trial, the UltraPro trial [
37]. The coefficient of variation of 0.3 in the number of patients between units is intended to allow for some variability between centers in the number of patients recruited. With an average number of 15 patients per cluster and per step, this corresponds to admitting that 95% of the clusters will recruit between 6 and 24 patients per step (between 60 and 240 in total). Correcting for 10% loss to follow-up, the same range will be between 67 and 267.
Of note, the 10% non-inferiority margin is supported by US FDA recommendations for anti-infectious trials assessing clinical success of a new treatment [
38]. Sample size calculation with a 5% margin would lead to 5100 patients jeopardizing the feasibility of the trial.
This sample size in the framework of a stepped-wedge design guarantees a power of more than 90% to prove superiority regarding antibiotics prescription (efficacy), if the proportion of antibiotics prescribed is 0.65 in the control group and 0.5 in the intervention group (α = 0.05).
The sample size was calculated with R software, based on the manuscripts by Hemming et al. [
39] and Harrison et al. [
40].
Statistical analyses
Datasets to be analysed, analysis populations
All analyses will be carried out in the primary analysis population, which includes all enrolled patients following an intention-to-treat principle. Any missing data at the enrolment visit due to incomplete documentation of the pre-assigned components of the algorithm will lead to the patient being assigned to the intention-to-treat analysis. Patients who are lost to follow-up will be replaced by additional patients’ inclusion as mentioned in the sample size (10% additional patients to compensate for loss to follow-up). Patients lost to follow-up will not be included in the analysis (complete case analysis). The primary efficacy and safety analyses will be repeated on the per-protocol population. The per-protocol population includes all patients who received all components of the PLUS algorithm without overruling on admission and 6–24 h later.
Stratified analyses will be done for severe LRTIs and non-severe LRTIs based on CRB-65 and PSI scores.
A sensitivity analysis will include patients who received a component of the intervention (PCT or LUS) while in the “usual care group” to evaluate if there is still an impact of the intervention when these tests are part of the usual care.
Additional analysis adjusting for patient level confounding factors (age, asthma, chronic obstructive pulmonary disease, CRB-65, and PSI score) in the intention-to-treat analysis will be done for both co-primary endpoints.
We will develop a detailed data analysis plan for secondary, subgroup, stratified, and sensitivity analyses. All analyses will be done with R Statistical Software.
Secondary analyses
The effect on bothersomeness of community-acquired pneumonia-related symptoms on day 7, day 28, and day 90 will be evaluated by comparing the number of points on the CAP symptom questionnaire between study groups by linear mixed effect models containing a random cluster level effect and a period-specific fixed effect. The effect on the duration of ED stay will be evaluated using survival models including frailty term for the cluster effect (Coxmodel).
Data safety monitoring board
The presence of such board was deemed unnecessary in view of the low risk associated with this study.
Time plan for the study
The planned overall study duration is 2 years and 3 months (113 weeks), including the recruitment period and study duration for each patient. Patient recruitment began in December 2022, the last-Participant-Out will be in March 2025.
Discussion
This clinical trial evaluates the safety and efficacy of a clinical algorithm including a clinical prediction score, PCT and LUS, to decide on antibiotic prescription in patients presenting with LRTIs to EDs. The clinical algorithm was built to optimize diagnostic performance to ensure safety and to maximize physician adherence to the algorithm’s recommendation.
The study population is selected using meaningful, reproducible, and nonrestrictive inclusion criteria, as well as only few exclusion criteria. In this pragmatic trial, the study population will be representative of the targeted real-life population. However, patients with severe underlying lung disease, severe immunosuppression, or clinical instability requiring ICU are excluded to ensure safety.
The stepped-wedge cluster randomized design of the trial overcomes many challenges faced during intervention studies targeting physicians and including patients with LRTIs. The cluster design prevents cross-contamination between study groups, as physicians are not exposed to the intervention during or before the control “usual care” period. Contamination is an important issue in trials when evaluating new diagnostic interventions focusing on physicians and where individual patients are managed by the same physicians, but randomized to different groups/managements. It is highly likely that diagnostic tools used in the intervention group spill over to the control group. The result of contamination is that outcome differences between the treatment groups would be diluted, biasing the trial towards the null hypothesis.
The stepped-wedge cluster design reduces the disadvantages and limitations of a parallel cluster randomized trial. It implies that all clusters (EDs) start in the control period. Then, the clusters switch to the intervention period at consecutive time points, where the time of the switch is randomized for every cluster. Eventually, all clusters will have switched from the control to the intervention period. The main advantage of this design is that the clusters act as their own controls because they are active both as control and intervention. Therefore, the intervention effect can be estimated from both between- and within-cluster comparisons. This results in more statistical power and smaller required sample sizes than in parallel cluster design [
44].
The stepped-wedge design also decreases the risk that the characteristics of the study population differ between groups, as it may be the case in a parallel cluster design, where EDs may admit patients with different characteristics. In addition, this design allows to control for and estimate the effect of seasonality on outcomes, since each season of the year and each year (controlling for yearly variability in circulating respiratory pathogens) will be represented in usual care and interventions groups. Therefore, this stepped-wedge study design is a type of cluster randomized controlled trial well-suited to study acute LRTIs and evaluating algorithm-based treatment decisions. It also facilitates post-study implementation in case of positive results, as all centers are exposed to the intervention at the end of the study [
45,
46].
This is a low-risk study for participants, comparable to standard of care, as it is based on previously validated tests. A potential risk is related to the inappropriate withdrawal of antibiotic treatment in patients with CAP which might be due to missing central infiltrates on LUS and to the non-perfect sensitivity of PCT. However, there are defined safety measures in the study design to diminish such risks: neither patients with severe symptoms, requiring intensive care, nor patients with severe chronic obstructive pulmonary disease (COPD) or severe immunodeficiency will be included. Moreover, procalcitonin-guided prescription in patients with LRTIs has been previously shown to be safe in patients with severe LRTIs [
20] in ED and in a primary care settings [
47,
48]. A clinical severity score will further ensure the safety of the intervention in those with discordant results between the components of the algorithm (LUS consolidation and low PCT). PCT will be repeated after 6–24 h in admitted patients who did not receive antibiotics. Patients discharged without antibiotics will benefit from a phone evaluation after 48 h.
In this pragmatic trial, we chose to compare our intervention to “usual care” instead of “standard of care” to confront it to real clinical practice. In order to increase homogeneity of the usual care, ED physicians of all centers will receive a training on the epidemiology and management of CAP in Switzerland based on Swiss guidelines [
26].
We expect that there will be some overruling, when antibiotic prescription is not recommended. To decrease the risk of overruling, we presented the rationale behind the algorithm during the site initiation visits. In patients in whom the algorithm is overruled and antibiotics are prescribed in spite of the recommendation of the algorithm, PCT will be repeated after 6–24 h. In case of low PCT levels, it will be recommended to stop antibiotics. A close monitoring will be performed to rapidly detect and correct non-compliance problems.
The PLUS algorithm based on clinical scores and easy-to-use diagnostic tests has the potential to improve the identification of patients with a LRTI who will benefit from antibiotics and reduce unnecessary antibiotic treatments. This diminished prescription of antibiotics will decrease side effects, cost, and resistance which are global health problems. The World Health Organization Trial Registration Data Set for PLUS-IS-LESS study is displayed in Table
7.
Table 7
World Health Organization Trial Registration Data Set for PLUS-IS-LESS study
Primary Registry and Trial Identifying Number | www.ClinicalTrials.gov, NCT05463406 |
Date of Registration in Primary Registry | 19 June 2022 |
Secondary Identifying Numbers | BASEC number 2022-00738 |
Trial Protocol version | Study protocol V3.0 dated 28.02.2023 |
Source(s) of Monetary or Material Support | Grant: SNSF 33IC30_201300 |
Primary Sponsor | CHUV |
Secondary Sponsor(s) | Not applicable |
Contact for Public Queries | NBB, noemie.boillat@chuv.ch |
Contact for Scientific Queries | NBB, noemie.boillat@chuv.ch |
Public Title | Procalcitonin and Lung Ultrasonography Guided Antibiotherapy in Emergency Departments, The PLUS-IS-LESS study |
Scientific Title | Procalcitonin and Lung UltraSonography based antibiotherapy in patients with Lower rESpiratory tract infection in Swiss Emergency Departments: a pragmatic stepped-wedge cluster-randomized trial |
Countries of Recruitment | Switzerland |
Health Condition(s) or Problem(s) Studied | Lower respiratory tract infections |
Intervention(s) | The PLUS clinical management algorithm: The PLUS algorithm starts with a validated pneumonia clinical prediction score (score of Van Vugt), followed by LUS. In case of positive results of any of these tests, PCT is measured to identify patients who will most likely benefit from antibiotics. A validated clinical severity score will ensure the safety of the intervention in those with discordant results (LUS consolidation and low PCT). Usual care: management as usual |
Key Inclusion and Exclusion Criteria | Inclusion Criteria: • Patients aged 18 years or more • Acute LRTI (acute illness, less than 21 days, with at least one lower respiratory tract symptom, i.e. cough, sputum, dyspnea, chest pain and no alternative explanation) • At least one of the following clinical criteria: ◦ Focal abnormal auscultation (decreased breath sounds, crackles, bronchial breath sounds) ◦ Fever (documented temperature ≥ 38°C in the last 24 hours, including self-measured temperature ≥ 38°C) ◦ Tachypnea (respiratory rate ≥ 22/minute) ◦ Tachycardia (heart rate ≥ 100/minute) Exclusion Criteria : • Previous receipt of a quinolone, macrolide or ceftriaxone or, of more than one dose of any other antibiotic within 72h prior to enrolment (excepted prophylactic antibiotics or antibiotics given for urinary tract infection) • Previous acute care hospital stay in the last 14 days • Cystic fibrosis • Severe COPD (≥GOLD 3 or if not available, as a proxy: exacerbation treated with antibiotics during the last 6 months) • Severe immunodeficiency (drug-induced neutropenia with <500 neutrophils/mm3, HIV infection with CD4<200 cells/mm3, solid organ or bone marrow transplant recipient, prednisone ≥ 20mg/day for >28 days) • Initial admission of the patient in the intensive care unit • Microbiologically-documented SARS-CoV-2 (in the last 10 days) • Lack of decision-making capacity |
Study Type | Type: Investigator-initiated, interventional, pragmatic study Allocation: Randomized Intervention model: Stepped-wedge rollout Masking: None (Open Label) Primary purpose: Diagnostic Phase: Phase IV |
Date of First Enrollment | 5 December 2022 |
Sample Size | 1530 patients |
Recruitment Status | Recruiting |
Primary Outcome(s) | Safety outcome: Proportion of patients with clinical failure at day 28 (defined as a composite of any of the following: death or secondary ICU admission or secondary admission to hospital or hospital re-admission after index hospital discharge or complications due to the LRTI [empyema, lung abscess]) Efficacy outcome: Proportion of patients prescribed an antibiotic in each intervention group between enrolment and day 28 |
Key Secondary Outcomes | Quality of life measured with the community-acquired pneumonia symptom questionnaire on day 7, 28 and 90 Duration of ED stay in each study group Rate and duration of hospitalization in each study group Proportion of patients prescribed an antibiotic in each study group between enrolment and day 28. Proportion of patients prescribed an antibiotic for a new respiratory infection in each study group between day 28 and 90. Antibiotic side effects and C. difficile infection, from day 0 to 28 Acceptability and feasibility of the intervention through extensive identification of barriers and facilitators in patients and physicians conducting qualitative semi-structured interviews Quality-adjusted life years (QALYs), derived from responses to the EQ-5D-5L questionnaire, in each group Cost of the intervention as compared to usual care Advanced automatic LUS image analysis method using machine learning to assist in LUS diagnosis and risk stratification. Proportion of physician reaching proficiency in LUS image/video acquisition and interpretation after the training module. Sensitivity, specificity and area under the ROC of physician “gestalt” and Van Vugt prediction score for CAP diagnosis |
Ethics Review | Approved on 29.11.2022 (BASEC number 2022-00738). |
Completion date | March 2025 |
Acknowledgements
We are grateful to all local physicians, their staff, and patients who participated and will participate in the study. Especially, we thank the staff of the emergency departments of the University hospital of Lausanne, University Hospital of Basel, Cantonal Hospital of Baden, Cantonal Hospital of St Gallen, Cantonal Hospital Baselland, Cantonal Hospital of Lucerne, St Clarasspital Basel, Hospital of Neuchâtel, Intercantonal Hospital of Broye, and Hospital Riviera-Chablais.
We also thank the study nurses of the study (Evelyne Arthemise, Pia Boesiger, Marie-Josée Brochu-Vez, Marie-France Derkenne, Sonia Eberhard, Hélène Gerhard-Donnet, Susanne Glienke, Constanze Herrmann, Jasmin Jentzsch, Cornelia Knapp, Sabrina Maier, Mathilde Quaireau-Rosa, Grazia Reinhard, Katrin Schmelzle, Antonia Sigrist and Barbara Sobhani) and the study monitors (Chahira Katamesh, Sara Mantero, Yuliya Plutino), the project managers (Aurélie Fayet, Maureen Redza-Dutordoir), and data managers (Vassili Soumas and Fady Fares).
The PLUS-IS-LESS study group consists of :
Steering committee: Principal investigators: Noémie Boillat-Blanco, Olivier Hugli, Werner Albrich, Roland Bingisser, Markus Schwendinger; External expert: Nicolas Garin; Trial statistician: Isabella Locatelli and Study coordinator: Cécile Bessat.
Lung ultrasound experts committee: Elena Garcia, Dominique Schwab, Yvan Fournier, Björn Mattsson, Stephan Gasser, Dieter Von Ow, Fabian Napieralski, Adriana Sirova.
Participating hospitals: University Hospital of Lausanne, University Hospital of Basel, Cantonal Hospital of Baden, Cantonal Hospital of Baselland, Cantonal Hospital of St Gallen, Cantonal Hospital of Lucerne, Intercantonal Hospital of Broye, Hospital of Neuchâtel, Hospital Riviera-Chablais, St Clarasspital Basel.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.