A new gene expression-based diagnostic blood test can separate bacterial and viral infections with 90% accuracy, the first to meet standards set by the World Health Organization.
In developing countries, most antibiotic prescriptions are not only pointless—an estimated 70% to 80% of them are given for viral infections, which the medications don’t treat—they’re also harmful, as overuse of antibiotics accelerates antibiotic resistance.
“Accurately diagnosing whether a patient has a bacterial or viral infection is one of the biggest global health challenges.”
A similar problem exists in the United States, where an estimated 30% to 50% of antibiotic prescriptions are given for viral infections.
The new test could allow doctors around the world to quickly and accurately distinguish between bacterial and viral infections, thereby cutting down on antibiotic overuse. The test is based on how the patient’s immune system responds to an infection.
It is the first such diagnostic test validated in diverse global populations—accounting for a wider range of bacterial infections—and the only one to meet the accuracy targets set by the World Health Organization and the Foundation for Innovative New Diagnostics to address antibiotic resistance.
Those targets include at least 90% sensitivity (correctly identifying true positives) and 80% specificity (correctly identifying true negatives) to distinguish bacterial and viral infections.
The new test is described in a paper in Cell Reports Medicine.
“Antimicrobial resistance is continuously rising, so there has been a lot of effort to reduce inappropriate antibiotic usage,” says senior author Purvesh Khatri, associate professor of medicine and biomedical data science at Stanford University. “Accurately diagnosing whether a patient has a bacterial or viral infection is one of the biggest global health challenges.”
Existing methods include growing the pathogen in a Petri dish, which takes several days, or polymerase chain reaction (PCR) testing, which requires knowing the specific pathogen to look for.
That’s why in many cases, “Doctors prescribe antibiotics empirically,” Khatri says. “They say, ‘We’re going to give you an antibiotic and if you get better, you had a bacterial infection. If you don’t, you have a viral infection, and we’ll stop the antibiotic.'”
Immune system reaction
The test is one of a new crop of diagnostic tests that look at the host response—that is, how the patient’s immune system is reacting—to identify the type of infection. They measure the expression of certain genes involved in the host’s immune response.
“The immune system has been doing this for millions of years, constantly learning what is bacteria, what is virus and how to respond to it,” Khatri says. “Instead of looking for the bug itself, we can ask the immune system.”
However, because these host-response tests have been designed using data from Western Europe and North America, they fail to account for the types of infections that are prevalent in low- and middle-income countries. In particular, they have trouble distinguishing the more subtle differences between intracellular bacterial infections and viral infections.
“Epidemiologically, bacterial infections in developed countries are usually from bacteria that replicate outside the human cell,” Khatri says. These extracellular bacteria include E. coli and those that cause strep throat. In developing countries, common bacterial infections like typhus and tuberculosis are caused by intracellular bacteria, which replicate inside human cells, as do viruses.
“The immune system has a different response based on whether it’s an extracellular or intracellular bacterial infection,” Khatri says. “The reason it gets tricky is because once the bacteria are inside the cell, the pathways overlap with the viral infection response.”
Current host-response tests can distinguish extracellular bacterial infections from viral infections with more than 80% accuracy, but they can identify only 40% to 70% of intracellular infections.
Point-of-care test next?
To develop a diagnostic test that can separate both types of bacterial infections from viral infections, Khatri’s team used publicly available gene expression data from 35 countries. These included 4,754 samples from people of various ages, sexes, and races with known infections. The diversity of patients, infections and types of data is more representative of the real world, Khatri says.
Using machine learning and half of these samples, they identified eight genes that are expressed differently in bacterial versus viral infections. They validated their eight-gene test on the remaining samples and more than 300 new samples collected from Nepal and Laos.
They found that these eight genes could distinguish intracellular and extracellular bacterial infections from viral infections with high accuracy, achieving 90% sensitivity and 90% specificity. It is the first diagnostic test to meet (and exceed) the standards proposed by the World Health Organization and the Foundation for Innovative New Diagnostics.
“We’ve shown that this eight-gene signature has higher accuracy and more generalizability for distinguishing bacterial and viral infections, irrespective of whether they are intracellular or extracellular, whether a patient is in a developed or developing country, a man or a woman, an infant or an 80-year-old,” Khatri says.
He hopes the new diagnostic test can eventually be translated into a point-of-care test and adopted by doctors in both developed and developing countries, as it requires only a blood sample and can be performed in 30 to 45 minutes. His team has applied for a patent on the test.
Additional researchers from Mahosot Hospital and the University of Health Sciences in Lao People’s Democratic Republic, Dhulikhel Hospital in Nepal, the University of Oxford, and the University of Toronto contributed to the study.
The Bill and Melinda Gates Foundation, the National Institute of Allergy and Infectious Diseases, the Department of Defense, the Ralph & Marian Falk Medical Research Trust, the Thomas C. and Joan M. Merigan Endowment at Stanford University, the Wellcome Trust of Great Britain, the National Science Foundation Graduate Research Fellowship, and the Stanford Graduate Fellowship funded the work.
Source: Stanford University
