Advice

Clinical and technical evidence

Clinical and technical evidence

A literature search was done for this briefing in accordance with the interim process and methods statement. This briefing includes the most relevant or best available published evidence which relates to the clinical effectiveness of the technology. Further information about how the evidence for this briefing was selected is available on request by contacting mibs@nice.org.uk.

Published evidence

The literature search identified 6 studies published in full that reported on the FebriDx test. The most relevant study to the NHS was a single-centre retrospective chart review in a UK GP practice (Davidson et al. 2017), which assessed the effect of FebriDx test results on therapeutic decisions. Of the other published studies, 1 was a prospective, single-centre, feasibility study to assess diagnostic accuracy (Sambursky and Shapiro 2015a), and 1 was a study in children only (Onrubia et al. 2020), and 2 (Self et al. 2017 and Shapiro et al 2018) were prospective, cross-sectional, observational studies done in the US comparing FebriDx with a reference standard algorithm that a physician could override. Clark et al. 2020 was a peer-reviewed study in people with suspected COVID‑19. A further 2 conference posters were identified; a multicentre feasibility study (Sambursky and Shapiro 2015b) and a single-centre feasibility study (Sambursky et al. 2014). The latter, single-centre feasibility study was not included for review because it included data which were duplicated in the fully published peer-reviewed study (Sambursky and Shapiro 2015a). An additional unpublished and non-peer reviewed study was identified for COVID‑19 (Karim et al. 2020).

The clinical evidence and its strengths and limitations is summarised in the overall assessment of the evidence.

Overall assessment of the evidence

Overall, there is adequate evidence in terms of quantity and quality for the FebriDx test. Of the 8 studies identified, 1 was an unpublished abstract, 6 were peer-reviewed journal articles, and 1 was an unpublished, non-peer reviewed study. In addition, a published economic study was identified, which is reported in the resource consequences section. Most of the studies had some involvement from the company either in design, financial or material support, introducing the potential for bias. Most studies were done outside the UK, predominantly in the US. However, there is clinical evidence in the UK, including for FebriDx's use in COVID‑19 testing, and economic evidence from a UK NHS perspective. There is evidence for the benefit of FebriDx in both children and adults. Evidence was found for its use in a real-world setting in people suspected to have COVID‑19 (Clark et al. 2020). Given the nature and timescale of the pandemic at the time of this review that evidence is understandably developing but limited. Karim et al. (2020) was identified in this area, this was unpublished and not peer reviewed, but is similarly supportive of the use of FebriDx in people with suspected COVID‑19.

Clark et al. (2020)

Intervention and comparator(s)

FebrixDx.

Reference standard: SARS-CoV-2 RNA detected by polymerase chain reaction (PCR) respiratory samples using QIAstat-Dx or laboratory PCR or both was considered as positive for COVID‑19.

Key outcomes

Sensitivity for detecting viral infection 93% (95% confidence interval [CI] 87% to 97%), specificity 86% (95% CI 79% to 92%). Overall accuracy 90% (95% CI 86% to 93%), positive predictive value (PPV) assuming a 20% prevalence of viral infection, 63% (95% CI 52% to 72% and negative predictive value (NPV) 99% (95% CI 96% to 99%). A predictive multi-variate model involving data from 201 people found that the addition of patient and clinical characteristics did not add significantly to the diagnostic accuracy compared with using the FebriDx MxA result alone.

Several patients with FebriDx viral positive results (negative by PCR) had radiological features of COVID‑19. So, these results were likely to be true positives despite negative PCR results.

Strengths and limitations

This is a relatively large real-world study. The reference standard is likely to be suboptimal, and the results cannot be applied to people who are immune-compromised or to children.

Onrubia and Gonzalez (2020)

Intervention and comparator(s)

FebrixDx compared with standalone C-reactive protein (CRP) testing.

Key outcomes

All 20 children were considered to have a bacterial infection using clinical assessment alone, and had point-of-care CRP testing and FebriDx. Standalone quantitative CRP testing showed a CRP of 20 mg per litre or more in 50% (10/20) of patients, who were deemed to have a bacterial infection.

FebriDx found that 10% (2/20) of patients had a bacterial infection and further differentiated 85% of patients as having a viral infection (17/20). Microbiologically unconfirmed respiratory illness was found in 5% (1/20) of patients. Both patients found to have a positive CRP line on FebriDx without an associated elevated myxovirus resistance protein A (MxA), also showed a quantitative CRP of 20 mg per litre or more.

Strengths and limitations

The study shows that FebriDx could improve diagnosis certainty and reduce unnecessary antibiotic use in children compared with both clinical assessment and CRP testing. The study involves a small sample of children from a single site.

Karim et al. (2020)

Intervention and comparator(s)

FebriDx compared with nasal and pharyngeal swab for viral PCR testing and an assessment by 2 physicians to determine the clinical likelihood of COVID‑19.

Key outcomes

There were 35 patients who had a positive FebriDx test result, all of whom had either positive PCR for COVID‑19 (30/35) or a clinical assessment suggesting COVID‑19 (5/35). This gave a PPV of 100%. For the 13 patients with viral negative test results PCR was also negative. Including 1 lower respiratory tract infection in which it was not possible to determine the exact cause of infection, and a viral infection could not be excluded, the NPV was 12/13 (92%). This excluded the NPV of PCR at 71% (12/17). Sensitivity was calculated at 100% for COVID‑19 (97% for viral infection) compared with 85.7% for PCR (COVID‑19). The specificity of both FebriDx and PCR for COVID‑19 was 100%.

Strengths and limitations

This study was a pre-print, which means it was not peer-reviewed or published. The study provides evidence of the technology's benefit in identifying COVID‑19. The evidence is mostly from adults and from a single centre.

Shapiro et al. (2018)

Study size, design and location

Prospective, cross-sectional, observational cohort study in the US of 223 people in emergency department and urgent care settings reporting a history of fever in the previous 72 hours who presented with clinical signs and symptoms of an upper respiratory tract infection.

Intervention and comparator(s)

FebrixDx compared with a reference method algorithm with physician override that included bacterial cell culture, respiratory PCR panels for viral and atypical pathogens, procalcitonin, and white blood cell count.

Key outcomes

One patient had an invalid FebriDx test, 2 patients did not have adequate testing for the reference standard and were excluded from the analysis. Sensitivity 95% (95% CI 77% to 100%), specificity 94% (95% CI 88% to 98%), PPV 76% (95% CI 59% to 87%) and NPV 99% (95% CI 93% to 100%).

For viral detection in patients reporting fever in the last 72 hours, when classifying results as viral or not viral, overall agreement was 87% (95% CI 82% to 91%). FebriDx showed a sensitivity of 90% (95% CI 83% to 94%), specificity of 76% (95% CI 66% to 84%), PPV of 83% (95% CI 77% to 87%), and NPV of 85% (95% CI 77% to 90%).

Strengths and limitations

Multicentre prospective study. No follow up after the index test, so it is difficult to assess FebriDX effect on antibiotic prescribing decisions. How study participants and samples were selected is described as convenience sampling, meaning it was determined by the availability of staff and the constraints of shipping biological samples. This may have biased the results.

Davidson et al. (2017)

Intervention and comparator(s)

FebriDx test results assessed to establish if they affected therapeutic decisions that would have been otherwise determined based solely on clinical exam findings. The patient's history and medical chart were followed up a month after the test to see if there were any subsequent medical consultations or hospital admissions.

Key outcomes

FebriDx altered clinical management in 48% (10/21) of patients and reduced antibiotic prescribing in 80% (8/10) of clinical cases of suspected bacterial infection.

Strengths and limitations

Study was done in a UK GP practice. The study used a small sample size. There was no inclusion or exclusion criteria reported.

Self et al. (2017)

Study size, design and location

Prospective, cross-sectional, observational study in the US of 371 people (206 people with suspected upper respiratory tract infection, 165 people who were part of an asymptomatic control group).

Intervention and comparator(s)

FebriDx compared with a reference standard for classifying upper respiratory track infection aetiology with an algorithm which could be overridden by 2 clinicians. There was also an asymptomatic control group (who did not have reference tests because the tests were presumed to be negative).

Key outcomes

All 206 people with suspected upper respiratory tract infection had a valid FebriDx test, but 1 person did not have adequate reference standard testing to enable a diagnosis, and so was excluded from the analysis. Two (1%) people from the asymptomatic control arm had invalid FebriDx tests.

FebriDx had overall agreement with the reference standard algorithm in 76.6% of cases. When classifying results as bacterial or not bacterial, overall agreement was 91.7%, sensitivity was 80%, specificity 93%, there was a PPV of 63% and an NPV of 97%. When classifying results as viral or not viral, overall agreement against the reference standard was 84%, sensitivity 87%, specificity 83%, there was a PPV of 64% and an NPV of 95%.

In the asymptomatic control group (163 people) specificity was 99%. FebriDx resulted in 2 false positive tests (1%); 1 false positive viral result and 1 false positive bacterial result.

Strengths and limitations

Multicentre prospective study. A convenience sample was used, which may have biased the results.

Sambursky and Shapiro (2015a)

Study size, design and location

Single-centre blinded clinical feasibility trial of 60 people in a hospital setting in the US. There were 12 people with suspected pharyngitis, 24 with suspected lower respiratory tract infection and an asymptomatic control group of 24 people.

Intervention and comparator(s)

Intervention: CRP and MxA-guided therapy with the FebriDx test.

Reference standard: clinical diagnostic algorithm with microbiology and laboratory assessments (PCR panels, bacterial cell cultures, enzyme-linked immunosorbent assay tests) and radiological assessment (chest X-ray).

Key outcomes

There were 2 invalid tests and 4 people were diagnostically indeterminate because of specimen leakage or rejection.

The FebriDx test correctly identified 92% (22/24) of patients as negative for infection, 80% (16/20) as having confirmed bacterial infection and 70% (7/10) as having confirmed viral infection.

The authors considered the FebriDx test to be a sensitive and specific method to differentiate acute febrile respiratory infections.

Strengths and limitations

Appropriate reference standards were used in the study.

The cohort was small and limited to adults (aged over 17 years). Therefore, results may not be generalisable to the younger population. No diagnostic accuracy outcomes (for example specificity and sensitivity) were reported. Control subjects were not clearly defined and were described as primarily admitted with suspected acute febrile respiratory infection.

Sambursky and Shapiro (2015b)

Study size, design and location

Prospective multicentre blinded clinical feasibility trial of 139 people in 11 medical institutions in the US. There were 56 people with confirmed infections, 81 with microbiologically unconfirmed respiratory illness and 2 people who were excluded because of incomplete data collection.

Intervention and comparator(s)

Intervention: FebriDx.

Reference standard: BioFire PCR respiratory panel, additional viral PCR tests, routine bacterial cell culture, procalcitonin, CRP, MxA, white blood cell count and Epstein-Barr virus immunoglobulin M and immunoglobuin G levels.

Key outcomes

In patients with confirmed bacterial infection, 95% (21/22) had CRP of 20 mg per litre or more. In patients with confirmed viral infection, 41% (14/34) had CRP of 20 mg per litre or more. Of these 14, the FebriDx test correctly identified 64% (9/14) of them as positive for viral infection.

It was reported that the FebriDx test would have reduced the over-prescription of antibiotics in 26% (9/34) of confirmed cases of viral infection compared with tests of CRP levels alone.

Strengths and limitations

The multicentre method gives a more generalised population. Consecutive enrolment reduces the potential for selection bias. Patient characteristics were not reported. A relatively large number of people were enrolled, but FebriDx test results were only reported for people with an infectious aetiology (41%, 56/137).

Sustainability

The company has not submitted any sustainability claims.

Recent and ongoing studies