Genoa team's two-protein blood test correctly sorted fever causes in all 85 babies tested

Fifteen bacterial infections among 85 babies is 17.6 per cent ^[1], close to the roughly 15 per cent Bellini gives for feverish infants generally ^[4]. So the sample resembles the babies emergency doctors see. Its limit is size. When a test goes 15 for 15, the exact 95 per cent confidence interval for its sensitivity, the share of bacterial infections it catches, runs from about 78 per cent to 100 ^[3]. The other direction is tighter. None of the 70 babies with a viral infection or no infection ^[2] was put in the bacterial group, and that places the lower bound on specificity near 95 per cent ^[4].

The two errors carry different costs. A false bacterial call gives a baby the precautionary antibiotics it would often receive today anyway ^[3]. A missed one leaves an infant with a bacterial infection untreated. Babies with fever "are considered high-risk until proven otherwise," Bellini said ^[2].

The planned trial is where that range can narrow. At a 15 per cent rate ^[4], 300 to 500 infants would include about 45 to 75 bacterial cases ^[5]. If the test again caught every one, the lower bound on sensitivity would rise to about 92 and 95 per cent respectively ^[6]. Running it at several hospitals ^[16] also checks whether the protein readings hold up outside a single team's hands.

The results so far come from a presentation at the European Emergency Medicine Congress in Paris ^[9]. The New Scientist account does not say how bacterial cases were confirmed, or whether the cut-offs for the two proteins were fixed before these samples were read. A cut-off chosen on the same 85 babies it is scored against has not yet been tried on a new baby.

The markers have a longer record. CD64 rises on neutrophils in response to bacteria and CD169 on monocytes in response to viruses ^[7], so the test reads one signal for each kind of pathogen. Fabrice Malergue of Beckman Coulter Life Sciences in Marseille, who has worked on the test for much of the past decade, reported in 2021 that the pair separates bacterial from viral infection in adults in an emergency room ^[11]. Other groups have since replicated that finding and shown the test works in children ^[12]. Infants were the open question. Malergue said that because their immune systems are still developing, some people doubted that the CD64 and CD169 markers could arise in them ^[13].

He says a collaborator using the test in Martinique during an outbreak of dengue, a viral and usually mild illness, found that some patients had leptospirosis, a bacterial infection that is life-threatening without antibiotics ^[15]. Malergue, who says testing babies in a hospital setting is a significant advance ^[14], is one of the test's developers. His group has built a benchtop machine that analyses finger-prick blood to call an infection viral or bacterial, and it is approved for research use ^[17].