Skip to main content

Personalised Risk assessment in febrile illness to Optimise Real-life Management across the European Union

Periodic Reporting for period 4 - PERFORM (Personalised Risk assessment in febrile illness to Optimise Real-life Management across the European Union)

Reporting period: 2020-07-01 to 2021-06-30

A major problem in the provision of paediatric services is detecting the small number of children who develop invasive bacterial infections among the many who present with the more common viral illnesses. A quarter of all children attending hospital emergency departments have a febrile illness but less than 5-10% of these turn out to be caused by an invasive bacterial disease. Current diagnostic methods rely on a culture of bacteria from blood, urine, or spinal fluid, but the results of cultures may take 48 hours or more to become available, too late to influence the decision to administer antibiotics.

Fear of “missing” life-threatening bacterial infection results in innumerable children with viral infections undergoing investigations, hospital admission, and treatment with intravenous antibiotics because the risk of bacterial infection cannot be excluded rapidly. Conversely, children with serious bacterial infections are “missed” and sent away from health care with the diagnosis of “viral infection” only to return critically ill with sepsis or meningitis. The financial burden to health care systems of investigation and treatment of children with fever is considerable, and the unnecessary use of antibiotics contributes to the global threat of antimicrobial resistance.

The PERFORM consortium aimed to apply new molecular and protein methods to identify biomarkers in the blood of children with fever to distinguish bacterial from viral infection. Rather than attempting to identify the causative bacteria, PERFORM aimed to identify the pattern of genes and proteins which change during the body’s response to infection, to identify “molecular signatures” distinguishing bacterial from viral infection. PERFORM studied febrile children attending hospitals across Europe, the Gambia, Nepal, and Taiwan, applying best current diagnostic approaches to identify those with microbiologically confirmed “definite” bacterial and viral infections. On these same samples, we used RNA sequencing and protein analysis linked to detailed biological and statistical analysis to identify a small number of human genes or proteins that together could be used as a “signature” to distinguish a unique response to either bacterial or viral infection. The most promising RNA and protein signatures were validated on independently diagnosed children and used to develop rapid tests that could be further validated prior to introduction into clinical use.
Since commencing PERFORM, over 6000 children have been recruited to the study in 9 European countries, the Gambia, Nepal, and Taiwan. Standardized definitions were used to assign patients to categories: “definite bacterial” infection (confirmed by bacterial culture from a normally sterile site), “definite viral” infection, “probable” bacterial or viral infections, or “uncertain” where a diagnosis could not be assigned with certainty.

In addition to current diagnostics, centralized molecular tests for a wide range of viruses and bacteria were undertaken in blood and throat swabs on over 3000 patients.

RNA sequencing of blood was undertaken on 2000 patients. The genes responding to infection were identified and novel statistical and bioinformatic methods were used to identify the smallest number of host RNA transcripts that distinguished bacterial and viral disease accurately.

A number of approaches were evaluated to rapidly detect RNA transcripts distinguishing bacterial and viral infection, including nanoparticle-based assays, and RT-qPCR assays. Partner BioMérieux used the FilmArray platform to detect the signature distinguishing bacterial from viral infection. A prototype FlimArray Pouch (Version 1) was designed, manufactured, and evaluated, demonstrating an accurate diagnosis of bacterial infection in over 2500 patients.

To identify a host protein biomarker signature distinguishing bacterial and viral infection, proteomic analysis by mass spectrometry was undertaken. Bioinformatic analysis of the data identified a protein signature distinguishing bacterial from viral infection, which was validated by immunoassay.

An observational study of 38000 children with fever attending emergency departments in European countries was undertaken to establish the patterns of febrile illness, identify clinical and vital sign variables that predict severe disease, and provide data for cost-effectiveness analysis of new tests.
• PERFORM successfully recruited, assigned diagnostic categories, and stored samples from over 7000 patients, with clinical information, providing a unique resource for biomarker discovery and evaluation.

• Current best practice diagnostic approaches, supplemented by centralized molecular tests for common bacteria and viruses, could only assign definite diagnoses in 24% of children (14% definite bacterial infection, 10 % definite viral infection). The remaining 76% of patients could not be assigned a definitive diagnosis.

• 86% of all recruited children (including those with viral infections) received antibiotics, highlighting the need for better diagnostics.

• Centralized molecular testing for bacterial and viral pathogens identified viruses in a high proportion of all patients (including 25% of those with bacterial infections). This surprising finding shows that detection of a virus does not exclude bacterial infection and suggests that bacterial and viral infections are not distinct entities but may interact to precipitate a severe illness. Pathogen-based detection of bacteria and viruses, even using molecular testing, does not clarify the diagnosis in the majority of febrile patients, emphasizing the need for new diagnostic approaches.

• Study of 38000 emergency department attendances for febrile children confirmed high rates of antibiotic usage, despite only a small proportion of patients with a definite diagnosis of bacterial infection. Analysis of this cohort has identified clinical features and vital sign-based algorithms that predict the likelihood of serious illness which can be used to identify patients requiring further evaluation, treatment, or admission.

• PERFORM has undertaken one of the largest blood RNA sequencing studies comparing the pattern of genes expressed in the blood of children with bacterial and viral infections. Analysis of this data has confirmed that bacterial infection can be distinguished from viral infection by the pattern of RNA genes expressed in blood. A “signature” based on a small number of RNA genes accurately distinguishes bacterial from viral infection and validates on independent data sets.

• Proteomic analysis and targeted immunoassay have identified novel protein biomarkers that accurately (>90%sensitivity and specificity) distinguish bacterial from viral infections and which were validated as diagnostic biomarkers on independent patient samples.

• PERFORM has developed a prototype diagnostic device to detect the RNA signature distinguishing bacterial from viral infection using partner BioMérieux’s FilmArray platform. In a large-scale validation study, the version 1 FilmArray pouch was applied to 3000 patient samples. Definite bacterial and viral infections were distinguished with high accuracy. Furthermore, the test classified 75% of patients without definite clinical diagnosis based on the probability of either bacterial or viral infection. The FilmArray RNA test can deliver a result in less than an hour from the time of blood sampling.
The PERFORM Clinical Network