Also in This Issue
What is diagnostic stewardship?
A tool that predicts imminent heart-attack risk
A simple way to diagnose oral cancer
Diagnosing Parkinson’s via skin biopsy
Rapid tests’ untapped infectious-disease potential
New and Noteworthy
Score high for imminent heart-attack risk? Maybe call your doc
If you were likely to have a heart attack in the next six months, wouldn’t you want to know? Scientists at Sweden’s University of Uppsala have developed a tool that estimates that risk, based on work they published in Nature Cardiovascular Research. Using data from across Western Europe, they were able to pinpoint a set of biomarkers whose values can predict the likelihood of an imminent heart attack pretty well (statistical evaluation of their model rated it as good, but not great). All of the biomarkers can be assessed using currently available blood tests.
The future belongs to AI, from the sublime to the ridiculous
Medical images are a richer source of clinically useful data than ever previously known, but AI is needed to enable human eyes to perceive much of it. This week, Eric Topol’s always excellent Ground Truths blog explains just how many previously unexploited diagnoses can be extracted from a simple chest X-ray, from cardiovascular risk, coronary artery disease, type 2 diabetes, even pancreatic cancer. One example: A case study published in the Lancet demonstrates that a human can easily miss seeing a significant lung tumor on a standard chest X-ray when the tumor is hidden behind other structures (the heart in this case), while AI can identify it.
That’s the sublime. Now for the ridiculous: A widely reported AI-based mobile app claims to be able to identify sexually transmitted disease status with 65-90% accuracy based only on photos of male genitalia. Clicking through to the scientific basis of this claim reveals only that it relies on H.O.P.E. (See image for their unique acronym.)
Commentary: It is unclear whether this is just dating-app spin (if he’s going to send the pic anyway, at least make it useful), since images like this have no value as a serious diagnostic (all STIs have long asymptomatic infectious delays). Either way, we just hope that no one is naïve enough to take these claims at face value.
A simpler way to diagnose oral cancer
A swab of an oral lesion found at the dentist or the ear, nose, and throat doctor can be used to diagnose cancer, according to a recent study in Cell Reports Medicine. The ratio of two proteins in the sample (human beta defensin 3 and human beta defensin 2), as compared to the same ratio in a swab taken from a normal area of the mouth, generates a score called the beta defensin index. Any score above a certain level suggests that the patient may have cancer and needs a biopsy. The study authors suggest that the test could reduce biopsies in primary-care settings by 95%.
Commentary: Dentists are diagnosticians who are often overlooked. In addition to cancer, they can diagnose diseases from diabetes (bleeding and swollen gums) to osteoporosis (bone recession) to GERD (enamel erosion) to HIV. All that said, while we are intrigued by the work done in this study, we won’t get excited until we see more data.
A skin biopsy test for Parkinson's disease
The standard diagnostic tools for Parkinson’s disease rely on “clinical consensus” symptomatic features (physical and cognitive). This technique has at least two major problems: The symptoms it looks for are also common to some very different neuropathies, and they manifest relatively late in the disease course.
A report in JAMA evaluates a very different approach to early Parkinson’s detection. Parkinson’s is the most prevalent of a group of neurodegenerative diseases called synucleinopathies, in which the protein alpha-synuclein builds up in nervous-system cells. The study looked at whether levels of phosphorylated alpha-synuclein (p-SYN) could be analyzed in a 3mm skin biopsy. When the test was performed on those with known synucleinopathies, sensitivity was high (best case: 99.1% with 85.3% specificity).
Commentary: This is yet another example of novel, less-invasive sample types proving capable of diagnosing complex disease. Unfortunately, this paper worked backwards from a pool of synucleinopathy patients diagnosed years earlier, and did not attempt to evaluate sensitivity for early onset Parkinson’s. (That would have taken decades of follow up, as was done in a recent Alzheimer’s study we reported on in March.) It also failed to collect CSF or plasma for comparison (as this 2022 PD CSF paper did). The field desperately needs quantitative diagnostic tools of this type - we have great hopes.
Food for Thought
The potential of lateral flow antigen tests for gonorrhea
One legacy of the COVID pandemic has been the public’s knowledge of and confidence (mostly) in lateral flow antigen (LFA) tests. These tests have been used for a variety of conditions - most widely used for pregnancy detection with urine as the sample.
The Lancet recently reported a South African study of a roughly $3 novel rapid antigen test for gonorrhea developed by FIND (a global health nonprofit) to meet WHO requirements for lower-income countries. The test proved to be highly accurate, with an 85-98% positive predictive value and a 93-98% negative predictive value (in a small sample of symptomatic patients).
Commentary: Molecular point-of-care tests dominate gonorrhea diagnosis in the developed world today, supported by their higher sensitivity, but that comes at 5-10 times the cost (see Nature’s The future of at-home molecular testing).The rapid antigen test has a lot of mileage in the infectious disease space, and as folks in the US take more personal control of their health we would like to see more progress in this direction. (Note: Mara is on the board of a company with related diagnostics.)
Diagnostic stewardship - a worthy goal with a long road ahead
You may have heard of antibiotic stewardship, the effort to combat overuse of antibiotics by limiting their use only to what is necessary and appropriate for a given condition. Now it has a counterpart: diagnostic stewardship. So what is that? A great overview of the concept appeared in a recent issue of the Journal of Hospital Medicine. The basic idea: “Better, more appropriate use of diagnostic testing can improve value and downstream care.” The improvements happen at three points in the diagnostics process:
When the test is ordered: Get “the right test for the right patient” by “nudging clinicians toward higher value tests or preventing unnecessary testing.” (In this context, “higher value” = the test that is most likely to yield actionable insights.)
When the test is being processed: Work toward improving test performance, analysis, and individualization.
When the test results are being read: “Improve how results are presented so that clinicians can better interpret and use results.”
Commentary: We agree with the authors’ stated reason for writing the article: “One driver of low-value care is the increasing abundance of, reliance on, and inappropriate use of low-value diagnostic testing.” However, separating the diagnostic wheat from the chaff, especially as new AI-powered tests appear on a daily basis, is a big challenge that’s going to require sustained systemic effort to achieve.
We reiterate: test value is not determined by test price or complexity - it is based on whether the test is the right one to use at that time. We also believe that physician education is critical here. As we reported in January, in a large study of 29 academic medical centers, 34% of diagnostic errors were caused by physicians failing to order appropriate tests.
Quick Hits
The FDA has authorized the first AI-based software for the diagnosis of sepsis. The authorization states that the software should only be used as an adjunctive tool and not as the only test for sepsis, and specifies that it be used on patients for whom a blood culture (the gold standard for sepsis diagnosis) has been ordered.
COVID EUA Update
The FDA issued no new COVID 510(k) premarket notifications, five new EUAs, one amendment to an existing EUA, and no new revocations in March 2024.
510(k) Premarket Notifications: 0
New EUAs: 5
Molecular (1): EZ-SARS-CoV-2 Real-Time RT-PCR
COVID Antigen (4): OSANG OHC COVID-19/Flu Antigen Test Pro Testo | CorDx Tyfast Flu A/B & COVID-19 Multiplex Rapid Test | SEKISUI Diagnostics OSOM Flu SARS-CoV-2 Combo Test | SEKISUI Diagnostics, LLC OSOM Flu SARS-CoV-2 Combo Home Test
Amendments to Existing EUAs: 1
Molecular: 1
Revocations: 0