Also in This Issue
Finding ctDNA in urine requires a new approach
A urine test to rule out aggressive prostate cancer
The present and future of smart pills
With AI explosion comes AI-journal explosion
Syndromic PCR beats culture when dxing pneumonia
New and Noteworthy
Prostate cancer urine test to rule out aggressive cases
Essentially all men who live long enough will eventually be faced with a high prostate-specific antigen (PSA) test result. The problem is that PSA is sensitive (95%) but non-specific (11%) for prostate cancer - that’s lots of false positives. So the next step is to be referred for MRI imaging, followed by biopsy.
The majority of high-PSA cases are caused by benign prostate enlargement. Of the remainder, the majority of cases are indolent, non-life threatening cancer that doesn’t require treatment. All this testing is scary, invasive, and costly. Yet even after all of it, the rate of unnecessary prostate removal remains unacceptably high: 40-60%.
A JAMA paper this week proposes a new 18-gene RNA PCR test that is highly effective at ruling out aggressive cases of prostate cancer. Called MPS2, the proposed biomarker panel retained 95% sensitivity for high-grade cancers but with a 95-99% negative predictive value.
Commentary: From a public health perspective PSA tests make little sense, but understandably individuals want to be sure that their case is not missed. There is clear evidence that when PSA screening rates go down, five years later prostate cancer deaths go up. So there is a real need for a test that reduces unnecessary invasive procedures without missing more aggressive cases. This test is designed to fill this important gap.
Urine ctDNA “liquid biopsies” require a new approach
The idea of being able to detect (non-renal, non-urinary tract) cancer using a urine test for circulating tumor DNA (ctDNA) is exciting, but it requires novel methods. The kidneys break up ctDNA into extremely small fragments of fewer than 50 base pairs (50bp). That’s much smaller than the 75 - 100bp chunks typically found in blood and plasma. This difference will likely present a challenge for all DNA tests using urine, including tests for pathogen DNA as a marker of infection.
A report from the University of Michigan (JCI Insight) described this problem, along with a prototype diagnostic test that can deal with it. The test was designed to detect 42bp disease-unique fragments and was created as part of an ongoing trial involving human papilloma virus-positive (HPV+) head and neck cancer. The trick, the team discovered, is finding places within the tumor/pathogen genome to target that are small enough so that the test can detect short strands without creating false positives.
Commentary: As previously discussed, the design of effective cancer tests requires deep and specific information about disease progression and the tissues involved. That is especially important if detection is to work during the disease’s earlier stages. Urine testing is yet another piece of evidence that one-size-fits-all detection does not work.
Smart pills are getting smarter
The first swallowable camera in a pill was introduced 25 years ago. Since then, the capabilities that can be fit inside a vitamin-size pill have grown exponentially, while cost and size have come down. Some of the innovations discussed in a recent Medscape overview of smart pill progress:
Sensors able to detect respiration and cardiac rhythms
Magnetically guided camera pills that can be controlled by a physician who can view images in real time
Coatings that can inform an app that a required drug dose has been swallowed
Universities in England and Scotland also recently shared their swallowable diagnostic, which has a new feature: It can detect how much pressure is being exerted in different areas of the gut and whether particular areas are expanding or contracting.
Commentary: Sensor miniaturization in both size and cost combined with AI-enabled analysis helps clinicians detect trends instead of isolated data points. There is much more to see (and swallow) in this field, and the future is near.
AI journals proliferate
A year ago, the prestigious New England Journal of Medicine (NEJM) announced the 2024 launch of NEJM AI, which has brought important insights and review papers to clinicians’ attention. The few existing back issues are definitely worth a read. Since then many other journals have been introduced and/or re-focused on AI:
AI in Medicine (ScienceDirect)
BMC Bioinformatics (an open-access publication from Springer Nature)
International Journal of Medical Informatics (ScienceDirect)
Two recent free-access review papers provide a useful basic introduction to the role of AI in early cancer detection (see graphic): A Primer on Artificial Intelligence and Exploring Clinical Implications.
Commentary: Given the crazy fast and broad growth of AI in medicine, we don’t think that the journals on this list even come close to saturating the market. However, what we need from all of them - and any publication that covers AI - is data from high-quality clinical trials and research studies without hyperbole. (Seriously. Not everything is “game-changing.”) We continue to believe that the first uses of AI will be in improving diagnosis in clinical medicine, automating physician processes, and accelerating drug discovery in research. Expect many more publications with AI in substance and name in the future.
Food for Thought
Syndromic PCR beats traditional culture for pneumonia dx
When a patient with suspected community-acquired pneumonia hits the ER, clinicians need a few critical pieces of information before they can provide the most accurate treatment. Is the infection viral or bacterial? If it’s bacterial, which bug is the cause, and what genes for resistance does it have?
Multiplex PCR panels called syndromic tests can provide the answer to those last two questions quickly, and they can do so even for bacteria that are hard to culture in the lab. But they can only find the genes they’re designed to look for. Good old traditional culture methods are slower, and fussier microbes can be tough to grow in lab conditions, but once you’ve grown a bug you’ve got it - even if it wasn’t what you thought you’d find. So which is better?
A study recently published in JAMA says that using a rapid (15-minute) PCR-based syndromic test is the way to go. It reported that there was no impact on initial antibiotic use (95% received some antibiotic within four hours), but patients in the molecular arm achieved pathogen-informed care 3x more often at the 48-hour point. Results were so convincing that the study was ended after interim analysis was complete.
Commentary: Why does it take an average of 30 hours for the results of a 15-minute test to impact patient treatment (versus 42 hours for traditional microbiology)? Part of the answer is the complicated logistics of using a centralized hospital laboratory (in this case, in Norway). Clearly early testing makes sense, but equally there is considerable room for improvement in timeliness.
Quick Hits
Lest you believe that every test needs to be high-tech, we present the results of a recent study of patients with bronchiectasis (a chronic condition in which the large airways of the lungs are abnormally widened). Clinicians were able to effectively predict the likelihood that a patient would get worse or even die, based only on “a four-point sputum color chart.” In other words, the more pus-filled a patient’s phlegm was, the worse off they were.