Review: In suspected influenza, some rapid tests have high sensitivity and high specificity for detecting infection
T. Fekete
DOI: https://doi.org/10.7326/ACPJC-2018-168-2-009
IF: 39.2
2018-01-16
Annals of Internal Medicine
Abstract:Question In children and adults with suspected influenza virus, what are the diagnostic accuracies of rapid influenza tests compared with reverse transcriptase polymerase chain reaction (RT-PCR) for detecting influenza A and B? Review scope Included studies compared commercialized rapid influenza tests (traditional rapid influenza diagnostic tests [RIDTs], digital immunoassays [DIAs], and rapid nucleic acid amplification tests [NAATs]) that provide results in 30 minutes, with RT-PCR as the reference standard for detecting influenza virus A or B in children or adults with clinically suspected influenza. Exclusion criteria included conference abstracts, casecontrol studies, use of reference standards based on index test results, or use of composite reference standards that included the rapid tests. Outcomes were sensitivity, specificity, and likelihood ratios (LRs). Review methods MEDLINE, EMBASE/Excerpta Medica, Cochrane Central Register of Controlled Trials, BIOSIS Previews, Scopus, and Web of Science (to May 2017) and reference lists were searched for peer-reviewed studies that were published in English, French, or Spanish and reported original data. 162 studies met the inclusion criteria: 130 assessed traditional RIDTs, 19 DIAs, and 13 rapid NAATs. Industry sponsored 20% of traditional RIDT studies, 68% of DIA studies, and 62% of rapid NAAT studies. 124 studies had sufficient data reported separately for influenza A and B and were included in the Bayesian bivariate random-effects meta-analysis. Main results Pooled test characteristics of each rapid test are shown in the Table. In prespecified subgroup analyses, respective pooled test sensitivities for influenza A and influenza B in children were 61% and 66% with RIDTs, 88% and 83% with DIAs, and 90% and 96% with rapid NAATs; in adults, pooled sensitivities were 43% and 33% with RIDTs, 75% and 57% with DIAs, and 87% and 76% with rapid NAATs. Pooled specificities were 96.7% to 99.9% over all tests. Conclusion In children and adults with suspected influenza virus, pooled sensitivity for influenza A or B was 92% to 95% with rapid nucleic acid amplification tests, 77% to 80% with digital immunoassays, and 53% to 54% with traditional rapid influenza diagnostic tests; specificities were >98%. Pooled test characteristics of rapid tests (reference standard RT-PCR) for diagnosing influenza A or B in children or adults with clinically suspected influenza* Influenza type Rapid tests Number of studies (n) Sensitivity (95% CrI) Specificity (CrI) LR+ LR A RIDT 94 (56399) 54% (49 to 60) 99.4% (99.1 to 99.7) 96 0.46 DIA 18 (10787) 80% (73 to 86) 98.3% (97.4 to 98.9) 47 0.20 Rapid NAAT 12 (2576) 92% (85 to 96) 99.2% (98.6 to 99.7) 116 0.08 B RIDT 30 (19073) 53% (42 to 64) 99.8% (99.7 to 99.9) 319 0.47 DIA 17 (10404) 77% (65 to 85) 98.7% (97.5 to 99.4) 60 0.23 Rapid NAAT 12 (2567) 95% (87 to 99) 99.4% (98.9 to 99.8) 166 0.05 *CrI = credible interval; DIA = digital immunoassay; NAAT = nucleic acid amplification test; RIDT = rapid influenza diagnostic test; RT-PCR = reverse transcriptase polymerase chain reaction; other abbreviations defined in Glossary. Commentary There are 2 reasons to seek diagnostic certainty for infections: clinical benefit and epidemiologic tracking. For influenza, it would be helpful both prognostically and therapeutically to confirm the diagnosis. Knowing that a patient has influenza will probably increase early use of antivirals and reduce reflexive ordering of antibacterials. But the public health benefits of tracking influenza are equally important, and rapid diagnosis permits timely prophylaxis of contacts, a push for vaccination, and possibly changes in policy around social interactions for vulnerable populations. Older rapid tests for influenza diagnosis have good specificity, but many clinicians do not order them because poor sensitivity may not justify the cost and inconvenience (1). In Merckx and colleagues' systematic review on newer diagnostic tests for influenza, the negative LR of up to 0.05 for the newer tests is exceptional and has the potential to affect care. Should we rethink our reluctance to order rapid tests for influenza? The answer depends on the reliability of the numbers in the review by Merckx and colleagues. There is some risk for bias based on industry sponsorship and uncertainty about patient selection. The rapid NAAT tests were not done at point-of-care and thus may represent different and probably optimized test conditions. The 2 most frequently evaluated rapid NAATs have somewhat different performance characteristics. Head-to-head comparisons of these molecular-based rapid influenza tests would be helpful. Ideally, we would generate outcome data based on availability of reliable influenza tests. However, even with these limitations, the newer tests are appealing enough to consider now if the price is right. In the old days, when cultures were done in many medical centers, granular data were available about influenza but not in time for clinical decision making and with some delay for public health. In the new world, we will still need details about precisely which strains are circulating, but we can track the movement by hemagglutinin and neuraminidase type. Who should pay for this? Some patients will be covered by insurance companies, including Medicare/Medicaid, and some costs will be absorbed by health entities (e.g., for inpatients). If we want public health organizations to support expanded testing, we need to fill the gaps in the meta-analysis by Merckx and colleagues and show that direct comparisons done by credible independent parties confirm the good results and that rapid NAAT tests perform well.