The validation study enrolled 1,222 patients (Liesenfeld O, et al. Nat Med. 2025;31[12]:4044–4054). The endpoints were clinically adjudicated infection status and need for critical care intervention within seven days. “Their inclusion criteria were suspected acute infection and one abnormal vital sign, or two abnormal vital signs and a blood culture ordered,” Dr. Farnsworth said. Like the other studies, the authors determined sepsis by expert panel adjudication. If two adjudicators agreed, it was considered a consensus decision (n=729); if not, a third was used. “They also did forced adjudication for part of their substudy. They ended up with 448 bacterial and 165 viral infections this way.”
In the very high TriVerity bacterial band score, the assay had a sensitivity of 35.9 percent and a specificity of 95.5 percent. In the very low group, it had a sensitivity of 97.2 percent and a specificity of 34.9 percent, “with the others falling in the middle between those two,” he said. In the very high group, the probability of infection was about 93 percent. In the very low group, bacterial infection was still possible—“in fact, 12 percent were still adjudicated to be bacterial”—but it was quite a bit lower than the other groups, even compared with the low group, which had a 48 percent probability of bacterial infection.
The viral scores had a similar breakdown. In the very high viral band score, sensitivity was 59.3 percent and specificity 98.6 percent. In the very low group, sensitivity was 95.5 percent and specificity 48.2 percent. Probability of viral infection in the very high group was 92.9 percent; it was 2.9 percent in the very low group.
Based on the results of the severity band, the authors assessed the probability that patients would require ICU-level care. “If you were in a high band, your probability of getting into the ICU was about 53 percent,” Dr. Farnsworth said. “If you were in the low band, it was two percent.” So it does help stratify patients by need for critical care intervention.
Those interested in viewing a comparison of these methods are likely to be disappointed.
“There has not been a true comparison in a real-world setting,” Dr. Farnsworth said. “And that makes sense. They’re different assays with different clinical indications.”
“That makes it a bit difficult for laboratories to know which technologies we should consider implementing,” he tells CAP TODAY.
Randomized controlled trials demonstrating clinical impact would help. “I think those are coming, and we have seen a few early studies emerge,” he says. Performance data from real-world studies, too, is needed. And it would be helpful if new sepsis guidelines included recommendations on how to implement emerging diagnostics in the ED.
He cites one study of 169 patients that compared the TriVerity and MeMed BV tests (Dedeoglu BE, et al. J Infect. 2024;89[6]:106360). “The caveat is that only four patients were admitted to critical care units and there were only two deaths. So this wasn’t a particularly sick population,” Dr. Farnsworth said. In some ways, that makes sense because it’s what the MeMed BV is intended for, he added, while TriVerity is more for the acute care population.
One thing is clear: Of all the new methods, there is no one best assay for the entire population.
Sepsis is a syndrome, first of all. “It’s very difficult for us to define,” he said. The clinical adjudication used in all these studies is helpful but has its limitations, “and is an imperfect gold standard.” And many of the studies exclude cases where consensus couldn’t be reached. “You don’t want super dirty data making its way into your study,” he said. “But at the same time, in real-life scenarios, it makes it difficult to know how something is going to perform.” Forced adjudication, on the other hand, can change study outcomes. “It’s hard for our providers to truly know what sepsis is, and when we force them to make a decision, it changes the characteristics of what we’re calling sensitivity and specificity,” he said. “It’s a limitation we have to be aware of.”
That’s why he sees it as a good thing that in some ongoing studies of emerging sepsis diagnostics, adjudicating the definition of sepsis has taken a backseat to other measures. Rather, the questions have become, “What happens in a hospital population if we implement this assay? Do we see improved outcomes and improved resource utilization?”
“That might be a better gold standard for us to go off than just sensitivity and specificity.”
Charna Albert is CAP TODAY senior editor.