AI mines primary-care data to diagnose cancer
Volume 11, Issue 20 | May 15, 2025
ALSO IN THIS ISSUE
FDA approves home sampling for HPV
Blood proteins allow ultra-early diagnosis of liver disease
Faster antibiotic selection with updated mass-spec
New sample type: You won’t believe your ears
Measles and Bird Flu Updates
FDA approves at-home sampling - an alternative to Pap smear
You may have seen a lot of headlines this past week about a newly FDA-approved at-home test for cervical cancer. We want to be clear - that’s not exactly what the agency approved, but what they did approve is still fantastic.
The device they green-lighted is a way for women to collect their own samples at home for human papilloma virus (HPV) testing. That may sound a lot less exciting, but since HPV causes almost all cases of cervical cancer, screening for the virus is actually better than the good old Pap smear at identifying women at risk of getting the disease.
The test requires a prescription, and the samples must be sent into a lab for analysis. It will be available to individuals ages 25 - 65 who are at average risk of cervical cancer.
AI mines primary-care data to diagnose hidden cancer
One of the initial hopes of AI was that it would be able to deeply analyze patients’ existing medical history to predict their medical future. For example, could an algorithm see patterns in tests taken / results received / patient age / family history that indicate the presence of undiagnosed cancer?
Researchers in the UK believe that they now have those algorithms, thanks to a study that involved more than 7 million patients. The National Health Service there already uses a scoring system called QCancer to look for patients at high risk of cancer. Researchers investigated whether new algorithms could improve on the accuracy of the QCancer score.
The new system includes more blood test results but also looks at seven symptoms (itching, bruising, back pain, hoarseness, flatulence, abdominal mass, dark urine). Compared to QCancer, it was about 5% more sensitive at discovering cancer in men and about 11% more sensitive in women. The authors also note that it is the first algorithm that can use data from primary-care clinicians to estimate the probability of having undiagnosed liver cancer.
Blood proteins allow ultra-early diagnosis of liver disease
One of the fastest-growing diseases around the world is the opaquely named Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD). The name translates to “liver with a lot of fat in it because of problems with the metabolism,” so it’s no surprise that the disease is often associated with Type 2 diabetes and obesity.
MASLD presents a diagnostic challenge. It causes few if any symptoms during the early stages, when lifestyle changes could potentially prevent things from getting worse. But if it progresses to its more severe inflammatory form, the liver gets irreversibly damaged.
Research presented earlier this month at Digestive Disease Week showed that when combined with traditional clinical risk scores, an increase in five proteins in the blood can predict an individual’s likelihood of getting MASLD. The model was able to predict disease five years before it began with 90% accuracy. Predictions 10 years in advance were 84% accurate, and those 16 years in advance were 81% accurate.
Guiding antibiotic selection with updated mass-spec
A perennial concern that never seems to get the attention it deserves is the use of ineffective antibiotics. A recent paper in Nature Communications shows how mass spectrometry (mass-spec) can be improved to aid antibiotic choice quickly.
Existing mass-spec methods look for proteins found only in bacteria. But these methods require isolating the bacteria from samples first. (See this early 2025 review – graphic below.) Instead of focusing on proteins from the bacteria themselves, the researchers behind the paper decided to look for substances the bacteria make during metabolism. That allowed them to use samples without extensive preparation. They also created a reference database of bacterial metabolic products. Using that tool resulted in greater specificity and sensitivity.
COMMENTARY: Mass-spec is quick (minutes, especially if sample prep is minimal) and low cost (in a central lab), but it’s relatively labor- and equipment-intensive. It also can only detect previously documented patterns in a reference database (i.e., it can only find things it already knows exist), and detecting emerging resistance is beyond its capabilities. However, there is no question that adding metabolites to the reference library of protein fragments can significantly increase clinical utility.
Bird Flu Update: Cases slowing - until the next migration
Concerning study results in cats, ferrets, and pigs
A global review of the last 20 years of research on bird flu in cats shows that infections in that species increased dramatically when clade 2.3.4.4b of the virus appeared. The strain is responsible for the current outbreaks and has a 53% mortality rate in domestic cats. Because cats aren’t monitored for bird flu, the authors suspect that the disease is underreported.
In better news, CDC officials noted that bird flu cases in both animals and humans have slowed down recently, likely because the spring bird migration season is over. They expect that cases will rise again in the fall.
Measles: US cases top 1,000
Measles appeared in wastewater two weeks before cases reported
More than 1,000 cases of measles have now been reported in the US this year.
Texas researchers detected measles virus in wastewater samples from the Houston area about two weeks before cases of the disease were reported in people. The results indicate that wastewater sampling could be used for early detection of measles.
The current larger outbreaks are shown below. The cases in Texas, New Mexico, Kansas, and Oklahoma all came from the same source; those in Arkansas, Indiana, Michigan, Montana, North Dakota and Ohio are unrelated.
You won’t believe your ears when we tell you about this sample
Here at Sensitive and Specific, we love to highlight innovative use of new sample types. Here’s one that never would have occurred to us: Earwax!
Scoring only slightly lower on the gross-o-meter than feces, earwax (its professional name is cerumen) has a lot going for it as a biological sample. It’s easy to access, can be collected non-invasively, and is made up primarily of lipids (fats). Blood is water-based, so the substances you find in it are things that dissolve in water. By contrast, earwax can tell you about what’s going on in the body with things that dissolve in oil or fat.
Researchers have already demonstrated that a set of biomarkers in earwax reliably indicates the presence of cancer, and have shown that earwax analysis can tell whether someone has Type 1 or Type 2 diabetes. Work is underway to look for earwax-based biomarkers of neurodegenerative diseases such as Parkinson’s and Alzheimer’s.
Thanks to one of our readers in Phoenix, Ariz., for highlighting this story for us. Keep those emails, cards, and letters coming!