Yes, superspreaders are super-shedders
Volume 8, Issue 3 | June 21, 2023
In This Issue
NGS enabled epis to link a deadly bug to eye drops
FDA pilots a way to regulate some LDTs
Diagnosing mpox with a smartphone and AI
The gut microbiome in early Alzheimer’s
New and Noteworthy
Superspreaders are super-shedders - and antigen tests can find them
We have known since the earliest pandemic days that superspreaders are responsible for the majority of COVID infections transmitted to others. What we did not know was whether these people were simply the friendliest among us (the ones who interacted with the most folks) or whether they were the unluckiest (those hit with a “perfect storm” of genetics, immune response, infection dynamics, and viral dose).
This week’s The Lancet Microbe finally gave us an answer, thanks to a COVID challenge study performed in the UK back in late 2020. Thirty-four human “guinea pigs” received a nasal spritz of SARS-CoV-2 under controlled conditions; 18 of them became COVID-positive. Two of those were responsible for 86% of all viable airborne virus collected, despite having only mild symptoms. And only 2% of the airborne virus circulated before antigen tests turned positive.
Sequencing was the key to solving the case of contaminated eye drops
When two cases of “extensively drug-resistant” Pseudomonas aeruginosa popped up in Los Angeles in late spring of 2022, they didn’t make sense to public-health workers. Both had a rare mutation that gives the microbe resistance to carbapenem-based antibiotics, and - even more strangely - both samples came from eyes.
In the end, it was the use of next-generation sequencing (NGS) by epidemiologists across the nation - and effective communication about those sequences - that cracked the case. A single scary strain of P. aeruginosa was ultimately traced to contaminated eye drops - and has caused four deaths, 14 cases of blindness, and 63 other serious eye infections so far.
Commentary: While PCR is unsurpassed in detecting known pathogens quickly and cheaply, NGS is the leader when either the causative pathogen is unknown, or as in this case, when a common genomic fingerprint connects multiple geographically separate cases.
FDA tests the waters for laboratory-developed test regulation (again).
This week the FDA announced a small pilot program that seeks to balance the speed with which laboratory-developed tests (LDTs) can be launched with FDA’s responsibility to make sure that diagnostic tests work as expected. The pilot addresses situations in which a given cancer drug has been FDA-approved, but the test that determines whether that drug is appropriate for a given patient has not been approved yet.
The FDA approach:
Get performance info for the tests that were used to enroll patients into the clinical trials for the drug in question.
Using those tests as a basis, set minimum performance characteristics for LDTs that will be used to determine whether patients can get the drug.
Commentary: This pilot is small (nine therapy/diagnostic combinations) and has a limited duration (one year), but we’re glad to see it. This approach might be a prototype for other diseases and companion diagnostics as well.
Food for Thought
Alzheimer’s diagnostics – can poop help?
A recent paper in Science Translational Medicine examined whether patients with no or minimal cognitive symptoms had detectable changes in their gut microbiome. The answer is promising. Substantial gut bacteria differences were found between those with tau/beta amyloid deposits and those without - specifically, high Bacteroidetes and low Firmicutes counts were typical of early cases. Interestingly, this pattern is typical of many inflammatory diseases, so poop tests may have limited specificity.
Commentary: Of course, correlation is not causation, but ways to identify the vulnerable at early stages are essential if we are to develop and test effective drugs. We are pleased to see the wide variety of Alzheimer’s tests in development - quite literally from top to bottom.
Diagnosing mpox with a smartphone and AI
Researchers in India tested the ability of five off-the-shelf trained deep-learning networks to diagnose mpox based on photos taken by smartphone. The networks were tested using a publicly available dataset which included both pictures of mpox lesions and skin lesions from similar diseases such as measles and chickenpox. The results, published in Medicine in Novel Technology and Devices, showed that the best network, ResNet-18, was able to attain an accuracy of 99.49%. Put in diagnostics terms, ResNet-18 reached 99.43% sensitivity and 100% specificity.
Commentary: And you thought tricorders would always be science fiction.
Are AI algorithms ready to make clinical judgements? Caution required.
Until now, clinical AI has been largely present interpreting images and cardiac electronic data. Now AI is taking a far more visible role in clinical assessment. On the one hand, AI can find patterns in complex data that might be invisible to human clinicians, but on the other, it is a black box how it draws its conclusions. This dilemma comes to a head with the challenge of diagnosing sepsis accurately and quickly - a recent discussion centered on Epic Systems’ algorithm for sepsis diagnosis. As reported in the Wall Street Journal, experienced nurses question the reliability of AI conclusions but feel unable to override them. The hospitals interviewed state the principle that AI is to be used to inform professional intervention before finalizing a diagnosis, but what this means in practice on a busy unit is hard to discern.
Commentary: At a minimum, AI enables discovery of patterns in clinical data at lightning speed. The challenge of AI: Is it accurate enough to rely on? Too early to tell - but we appreciate the practical attempts to combine Artificial and Human Intelligence but exactly how to do this in practice is unclear.
Quick Hits
The National Wastewater Surveillance System, which currently tracks only SARS-CoV-2, is set to expand. CDC plans to add tracking for influenza, RSV, E. coli, norovirus, mpox, and other other pathogens to the system, as well as genes for antimicrobial resistance.