Can a protein test distinguish bacterial from viral infections when doctors are unsure?

Details

Viral and bacterial illnesses often look similar, making the cause of infection difficult to determine. MMBV is a host-protein test that computationally combines circulating levels of TNF-related apoptosis-induced ligand, interferon γ-induced protein-10, and C-reactive protein to distinguish bacterial from viral infection. It had shown high performance in multiple prospective clinical validation studies. This study evaluated its diagnostic accuracy in febrile children whose physicians were uncertain about the cause, when the test was used at the physician’s discretion.

Patients aged 3 months to 18 years were retrospectively recruited for the SPIRIT study (NCT03075111; 2014–2017). At blood draw, emergency department physicians recorded their suspected cause and level of certainty in a questionnaire. Prespecified MMBV scores indicated viral/non-bacterial infection at 0 ≤ score <35, an equivocal result at 35 ≤ score ≤65, and bacterial infection or coinfection at 65 < score ≤100.

Three independent experts adjudicated the reference-standard cause as bacterial, viral, or indeterminate using all available patient data. They were blinded to the MMBV results. Both MMBV and physicians’ suspicions were assessed against this reference standard.

Of 3,003 potentially eligible patients, physicians were uncertain about the infection’s cause in 736 cases assigned a reference standard: 128 bacterial and 608 viral. MMBV sensitivity was 89.7% (96/107; 95% confidence interval 82.4–94.3) and specificity was 92.6% (498/538; 95% confidence interval 90.0–94.5). It significantly outperformed physicians’ suspicions, which had sensitivity of 66.2% (49/74) and specificity of 72.0% (265/368; P < .0001). MMBV results were equivocal in 12.4% (91/736) of cases.

The authors concluded that MMBV was more accurate than physicians’ suspicions in determining the cause and had high sensitivity and specificity against the reference standard.