Today's stories ask how far to trust a result: a strong heart trial with a design flaw, new aging biomarkers, and a contested vaccine-injury effort.
Picked from 33 new articles across 12 sources · arrow keys for other days
Today in 30 seconds
30s left
Tricuspid clip repair cut heart-failure admissions about 65% in 360 elderly Germans, but the open-label design may inflate it. story 1 ↓
A memory T-cell split, measured as a blood protein, predicted 17-year mortality in 48,620 UK Biobank participants. story 2 ↓
Clinicians' diagnostic accuracy rose from 82% to 93% when lab results showed likelihood ratios instead of normal/abnormal flags. story 3 ↓
HHS opened an NIH vaccine injury clinic October 5 with no director or budget named; critics say it's designed to find harm. story 4 ↓
In mice, boosting chaperone-mediated autophagy helped immune cells clear senescent cells and eased lung fibrosis. No human data yet. story 5 ↓
1/5Sensible Medicine· John Mandrola· Oct 12
Tricuspid valve repair trial looks strong, but a cardiologist sees bias
A German trial found that clipping a leaky tricuspid heart valve sharply cut heart-failure hospital admissions. Cardiologist John Mandrola argues that unblinded design, crossover rules and incomplete follow-up make that result unreliable.
Why it matters · Millions of older adults have a leaky tricuspid valve. Whether clip repair should become standard care depends on whether its benefit is real or comes from how the trial was run.
Know this first
Context from the editor, not the article · 28 sec
1 · The 'forgotten' heart valve
The tricuspid valve sits between the right atrium and right ventricle. When it leaks (tricuspid regurgitation), blood backs up into the body, causing abdominal fullness and leg swelling. It becomes much more common with age.
2 · Why TR is hard to treat
The leak is usually driven by high pressure coming from the left side of the heart, and it changes a lot from day to day: after diuretics lower the pressure, severe leakage can shrink substantially. Valve surgery alone carries high complication rates.
3 · Catheter fixes and earlier trial
2/5Ground Truths (Eric Topol)· Eric Topol· Oct 11
One kind of memory T cell tracks unhealthy immune aging and earlier death
Researchers sequenced 12.4 million single blood cells and found aging T cells take one of two paths. A blood-protein estimate of the "granzyme B" path predicted higher death rates in UK Biobank over 17 years. The data are observational.
Why it matters · Doctors have no routine way to measure how well a person's immune system is aging. This work points to a cheap blood-protein test, though it hasn't been shown that changing the measure changes anyone's risk.
Know this first
Context from the editor, not the article · 28 sec
1 · The thymus and naive T cells
The thymus trains new T cells. It starts shrinking early in life and shrinks faster after puberty. Over the decades the supply of fresh, "naive" T cells falls, and memory T cells, which have already met infections, fill the space.
2 · Granzymes
Granzymes are enzymes stored in packets called granules inside killer T cells. Granzyme B, released with perforin, makes infected or cancerous cells die. Granzyme K kills less well and has been linked to inflammation.
3 · Inflammaging
Inflammaging is the chronic, low-grade inflammation that rises with age. It is linked to heart disease, diabetes and frailty. Blood markers such as C-reactive protein (CRP) give a rough reading of it.
3/5The Conversation: Health· Amiran Baduashvili, Associate Professor of Medicine, University of Colorado Anschutz· Oct 12
Doctors made better calls when lab results came with likelihood ratios
A physician-educator argues that "normal" and "abnormal" lab labels hide crucial information. In his team's simulated-case trial, clinicians shown how strongly each value points toward or away from a diagnosis chose correctly 93% of the time, versus 82%.
Why it matters · Anyone reading results in a patient portal sees the same binary flags that misled the doctor here, and knowing what a specific number means can change care.
Know this first
Context from the editor, not the article · 27 sec
1 · Reference ranges and cutoffs
Labs flag results using a cutoff, often set from the spread of values in healthy people or from a diagnostic threshold. Results are labeled "normal" or "abnormal" no matter how close to the line they fall.
2 · What a likelihood ratio is
A likelihood ratio compares how often a given result shows up in people who have a disease versus people who don't. Above 1 makes the disease more likely, below 1 less likely. Near 1 changes little.
3 · Patients see results first
4/5Science-Based Medicine· David Gorski· Oct 12
HHS opens NIH vaccine injury clinic and plans to pay doctors for injury reports
The US health department opened a vaccine injury clinic at the National Institutes of Health and announced paid injury reporting and record changes. David Gorski argues the effort is designed to inflate apparent vaccine harms.
Why it matters · These changes affect the main US systems for spotting vaccine harms, and those systems' data will shape future decisions about which vaccines stay on the schedule.
Know this first
Context from the editor, not the article · 30 sec
1 · VAERS: passive reporting
The Vaccine Adverse Event Reporting System accepts reports from anyone about any health problem after vaccination. It is an early-warning tool. A report does not show the vaccine caused the problem, and any signal it raises must be confirmed by stronger studies.
2 · Active surveillance exists too
The Vaccine Safety Datalink, run by the CDC with large health systems (mostly Kaiser Permanente), automatically scans patient records for events that cluster after vaccination. It does not depend on anyone choosing to file a report.
3 · Who runs HHS now
Robert F. Kennedy Jr., a longtime vaccine critic, heads the Department of Health and Human Services. Last year he fired the CDC's vaccine advisory committee. Jay Bhattacharya runs the NIH, and Mehmet Oz runs Medicare and Medicaid.
5/5Fight Aging!· Reason· Oct 12
Aging cells' failing protein disposal may stop immune cells clearing them
A mouse study finds that as a selective protein-disposal pathway declines with age, senescent "zombie" cells send signals that blunt the immune cells meant to clear them. Restoring the pathway with a drug reduced senescent cells.
Why it matters · It suggests a second route to clearing senescent cells: help the immune system remove them, rather than killing them with senolytic drugs. So far the evidence comes only from mice.
Know this first
Context from the editor, not the article · 26 sec
1 · What senescent cells are
Cells that permanently stop dividing after damage, stress or too many divisions. They release inflammatory signals. In youth, immune cells clear them quickly. With age they build up and are linked to tissue dysfunction and disease.
2 · Autophagy in brief
Autophagy covers the ways cells break down their own worn-out parts inside lysosomes, small acid-filled compartments that act as recycling bins. Several forms exist. Some engulf large chunks of the cell, while others pick out individual proteins.
3 · Macrophages
Macrophages are innate immune cells that engulf debris, pathogens and unwanted cells. They are one of the main ways the body removes senescent cells in healthy young tissue.
Two catheter devices are FDA-cleared: a clip that pins leaflets together, and the EVOQUE replacement valve. The earlier TRILUMINATE clip trial improved quality of life but did not reduce deaths or hospitalizations.
tricuspid regurgitation (TR)
Backward leakage through the valve between the right atrium and right ventricle, causing fluid buildup in the body.
Image: Sensible Medicine
The news
TRIC-I-HF, a German trial of 360 elderly patients with a leaking tricuspid valve, found that catheter-based clip repair sharply cut hospital admissions for heart failure compared with medical therapy alone. The authors and the accompanying editorial called it a clear win. Cardiologist John Mandrola, writing at Sensible Medicine, argues that the trial's design leaves too much room for bias to trust it yet.
The problem
Tricuspid regurgitation (TR) is a leak in the valve between the heart's right chambers. It causes abdominal fullness and heavy leg swelling. It is usually driven by pressure from the left side of the heart and changes a lot with fluid status. Surgery on this valve alone is risky. Catheter devices threaded up from a groin vein now offer another option. The one used here, 1, clips the valve leaflets together.
The trial and its numbers
Patients averaged about 80 years old and were very sick, with 17.5% one-year mortality in the control arm. The trial was 2, so everyone knew who had the procedure. It randomized 2:1, and 98% of repair patients got TEER.
3 of 2.42. This method compares every repair patient with every control patient, first on survival, then on heart-failure hospitalizations, then on quality of life.
Free of death or heart-failure admission at three years: 52.4% vs 21.0%.
Heart-failure hospitalizations: 19.4% vs 43.9%.
Deaths: 13.1% vs 16.3% at one year (not significant), and 19.0% vs 27.6% at three years.
Mandrola's objections
Admissions drove the result. Hospitalization accounted for 9,280 repair wins against 2,871 for control. Death decided only a small share of pairs. Admitting a patient is a judgment call, and doctors knew who had the device.
Crossover needed an admission. Control patients could get the procedure only after a heart-failure hospitalization. 39% crossed over, most during that same stay.
Other measures didn't match. Weight, swelling, walking distance and quality of life barely improved, which is hard to square with a 65% drop in admissions.
Repair was partial. Only 70% of repaired patients had moderate or less leakage at 30 days.
Three-year deaths are premature. Only 20% of patients had reached three years of follow-up.
Baseline imbalances. The control arm was somewhat sicker.
What would settle it
Mandrola wants a 4 trial, in which control patients undergo a mock procedure so they don't know which group they're in, as has been done for coronary stents. He also wants cardiovascular deaths reported separately from all deaths. He refers patients for this procedure and hopes it works, but says this trial can't show that it does.
The news
TRIC-I-HF, a German trial run by investigators rather than a manufacturer, tested catheter-based repair of the tricuspid valve against medical therapy alone. On paper it is the strongest result yet for treating a leaky tricuspid valve. Repair cut heart-failure hospitalizations by roughly 65%, and three-year deaths were lower. The authors and the accompanying editorial called it a win. Writing at Sensible Medicine, cardiologist John Mandrola counts at least six reasons to hold back, and most of them come from the trial's design.
The problem being treated
Tricuspid regurgitation (TR) is a leak in the valve between the right atrium and right ventricle. Patients get a full, swollen abdomen and heavy leg swelling. It usually isn't a problem with the valve itself. Pressure overload from the left side of the heart stretches the right side, though pacemaker leads can sometimes pin a leaflet open. TR is also "highly dynamic": a patient admitted with severe TR who gets diuretics (drugs that make the kidneys shed fluid) may have much less leakage a few days later. That makes it hard to judge how bad the disease really is, and how much any fix helps.
For years the only options were adjusting diuretic doses or surgery, and surgery on the tricuspid valve alone has had very high complication rates. Catheter approaches changed that. The valve is a straight path from a vein in the groin. Two devices are FDA-cleared. One does 1, which clips the leaflets together. The other is the EVOQUE valve, a full replacement. The earlier TRILUMINATE trial of the clip compared the procedure with pills and no sham procedure. Its benefit came entirely from quality-of-life scores, with no fewer deaths or heart-failure admissions. Mandrola calls it one of the most biased trials in cardiology, because patients who know they had a procedure tend to report feeling better.
How TRIC-I-HF was designed
Results
4: 2.42 (95% CI 1.76–3.33).
Freedom from death or heart-failure hospitalization at three years: ( 0.40, 95% CI 0.29–0.55). A hazard ratio of 0.40 means events happened at about 40% of the control rate.
Mandrola's six objections
1. The result rests on a subjective endpoint. In the win ratio, heart-failure hospitalization produced 9,280 wins for repair against 2,871 for control. Death contributed far less (4,366 vs 3,166), and 74% of pairs tied on death. Admitting a patient is a clinician's decision. When the doctor knows who has the device, Mandrola argues, it is easy to increase oral diuretics for a repaired patient and admit an unrepaired one.
2. The crossover rule rewarded admissions. Control patients could cross over to repair only after a heart-failure hospitalization. The procedure is reimbursed and available in Germany, so a control patient or doctor who wanted it needed a qualifying admission. 48 of 123 controls (39%) crossed over, and 28 of those 48 (58%) had the procedure during that same hospital stay.
3. Congestion measures barely moved. If the repair cut fluid overload enough to prevent 65% of admissions, you would expect clear gains in body weight, swelling and six-minute walk distance. None were impressive. The 6 (Kansas City Cardiomyopathy Questionnaire, a heart-failure quality-of-life score) improved only 6.9 points more with repair, less than in TRILUMINATE. Earlier trials showed large quality-of-life gains and no effect on admissions. This one shows the reverse. Mandrola argues the two should move together if the benefit is real.
4. The leak was only partly fixed. At 30 days, 70% of repair patients had moderate or less TR, compared with 30% of controls. That is a less complete repair than in TRILUMINATE, which makes the large drop in hospitalizations harder to explain. A threshold effect is possible but unproven.
5. The three-year death data are thin. Supporters point out that death is an unbiased endpoint. But only 20% of patients had completed three-year follow-up. At one year, when follow-up was complete, the difference in deaths was not significant.
6. Baseline imbalances favored the device arm. The control arm had somewhat more patients with a higher TRI score (a surgical risk score for tricuspid disease), atrial fibrillation, coronary artery disease, prior heart attack and stroke.
What would settle it
Mandrola doesn't suspect bad intent. He says device trials are inherently harder than drug trials, where a placebo provides blinding. His minimum standard is a 7 trial, in which control patients go through a mock procedure so they don't know which arm they are in. He notes this has been done for coronary stenting. A clip is visible on X-ray, so doctors could still tell, but patients would be blinded. He also wants deaths split into cardiovascular and all-cause, because 80-year-olds die of many things unrelated to the valve.
Mandrola is not dismissing the procedure. He has referred patients with severe TR who were often turned down for anatomical reasons, and he says a working repair would help many people. His view is that this trial cannot show whether it works.
Remember
In open-label device trials, clinician-decided endpoints like hospitalization are vulnerable to bias. Death is not.
Repair cut admissions by about 65% but barely improved weight, swelling or quality of life, an inconsistency Mandrola flags.
The three-year mortality benefit rests on only 20% of patients with complete follow-up; one-year deaths didn't differ significantly.
Topol's AI-made schematic of two memory T cell types with different granules and different effects.1 / 4
The finding
A study published October 9 in the journal Immunity, led by Maxim Artyomov, mapped how white blood cells change across adult life. It found a fork in the road for aging T cells, and the branch a person's cells take is linked to mortality. Cardiologist Eric Topol, writing in Ground Truths, calls it one of the most important aging papers he has seen.
What happens as T cells age
The thymus, the gland that trains new T cells, shrinks with age. So the supply of fresh "naive" T cells and 1 (T cells that have only just left the thymus) falls steadily. In this study, that decline was the best single predictor of a person's age in years.
What matters more, Topol argues, is what replaces them. These are 2 (Tem), "killer" cells that remember past infections. The researchers found these cells split into two kinds:
TemK+ cells, whose internal granules mainly hold the enzyme granzyme K.
TemB+ cells, which mainly hold 3.
In mouse experiments, TemB+ cells were exhausted and "leaky." They spilled their contents, damaging tissue, activating macrophages and stoking inflammation. Two things pushed cells down the B path: past infection with 4, a common lifelong herpesvirus, and chronic heart or metabolic disease such as type 2 diabetes.
The split only showed up because the team used 5, which reads gene activity cell by cell. They did this for 12.4 million cells from 2,609 people aged 18 to 98 in 8 cohorts. Older methods that sort cells or average across them had missed it.
From cells to a blood test
Single-cell sequencing is too costly for a clinic. So the team used 6, an assay that measures thousands of blood proteins, to build a protein signature that estimates the share of TemB+ cells. It closely matched the single-cell data in 4 cohorts.
In the 7, a large British health study, they applied it to 48,620 people followed for 17 years. The third of people with the highest TemB+ estimate had markedly higher all-cause mortality. High TemB+ also predicted several specific diseases, and it was higher in people with CMV infection or high C-reactive protein (CRP), a blood marker of inflammation. It did not predict Alzheimer's, Parkinson's, dementia or macular degeneration.
The article shows these results only as charts and gives no hazard ratios, so the size of the effect is unclear.
Caveats
The data are observational. Nothing shows TemB+ cells cause harm, or that lowering them helps.
Each person was measured once.
The cohorts were mostly of European and Asian ancestry.
Only blood was studied, not tissue.
UK Biobank lacked CD8+ T cell counts, so these had to be estimated with a model.
"Good K, bad B" oversimplifies, because granzyme K can also drive inflammation.
Topol notes that the researchers are developing a cheap test. Whether such a test should change care is still unproven.
The news
On October 9, 2026, the journal Immunity published a study from a team led by immunologist Maxim Artyomov. It is the largest single-cell survey yet of how human white blood cells change with age. Its main finding: as the body runs low on fresh T cells, the replacement cells follow one of two paths, and the path a person's cells take is linked to how long they live. In his newsletter Ground Truths, cardiologist Eric Topol calls it "one of the most important papers in the science of aging," though he also calls its title one of the worst he has seen.
The problem: an immune system we can't measure
Topol makes a point he has argued before. Medicine has many ways to change the immune system, including checkpoint inhibitors that release its brakes in cancer, B-cell depletion, and so-called inverse vaccines that train it to tolerate the body's own tissue in autoimmune disease. Yet no routine clinical test tells you how a person's immune system is doing. "We can dial it up or down but can't measure it!"
Part of the reason is technical. Older methods could not reliably separate the relevant subtypes of 1 (Tem). These are "killer" T cells that remember past infections. Flow cytometry sorts cells by surface markers, and pooled (bulk) RNA sequencing averages the signal across millions of cells. Neither method could tell these subtypes apart. Single-cell RNA sequencing (scRNA-seq) reads which genes are switched on in each cell separately, and run at scale it could.
Phase 1: 12 million cells
The team combined data from 8 cohorts around the world: 2,609 people aged 18 to 98, and 12,394,834 white blood cells, grouped into 59 immune cell types.
The single strongest predictor of a person's age in years was the loss of naive T cells and 2 (RTEs), which are T cells that have only just left the thymus. That matches what is known about the thymus shrinking with age. Topol stresses that the loss of these "rookies" matters less than what replaces them, the "veterans."
The veterans are CD8+ effector memory T cells, and the study found they split into two kinds:
TemK+ (GZMK+): memory cells whose granules mainly hold granzyme K.
TemB+ (GZMB+): more fully matured ("terminally differentiated") cells whose granules mainly hold 3.
In mouse experiments, the TemB+ cells were "leaky" and exhausted. They spill their granules, which breaks down the supporting scaffold between cells (the extracellular matrix), damages tissue, activates macrophages and drives inflammation. Topol's shorthand is that B stands for "bad." Two things pushed cells toward the B path: past infection with 4, a common herpesvirus that stays in the body for life, and chronic disease, especially heart disease and metabolic disease such as type 2 diabetes.
An analogy (Topol's framing, extended): when a team loses its rookie pipeline, it fills the roster with veterans. Some are steady players. Others are worn out and keep committing fouls that hurt their own side. The share of fouling veterans matters more than the headcount.
Phase 2: turning it into a blood test
Running scRNA-seq on millions of cells is too expensive to use as a clinical test. So the researchers turned to 5, a commercial assay that measures thousands of proteins in blood plasma at once, in 4 of the 8 cohorts. They used 6, a statistical method that picks out a small set of useful predictors from many candidates, to find a protein signature that estimates the percentage of TemB+ cells. In each of the 4 cohorts, the protein-based estimate closely matched the single-cell measurement.
They then tested the protein signature in the 7, a large British health study: 48,620 participants followed for 17 years. They split people into thirds by estimated TemB+ percentage. The top third had clearly higher all-cause mortality than the middle and bottom thirds, and the curves kept moving apart over the follow-up. High TemB+ also predicted a range of specific diseases.
The article does not report hazard ratios or absolute risk differences. It shows these results only in figures, so the size of the effect can't be given here.
Two findings strengthen the biological story:
People who tested positive for CMV antibodies had higher GZMB+ levels.
People with higher CRP had a higher TemB+ percentage.
One finding sets a limit on it. TemB+ did not predict Alzheimer's disease, Parkinson's disease, dementia or age-related macular degeneration. Whatever this measure captures, it does not appear to track brain and eye degeneration.
Caveats
Topol lists several, and they matter:
Observational, not causal. Nothing here shows that TemB+ cells cause the deaths or diseases, or that lowering them would help.
One measurement per person. Nobody knows how TemB+ changes within a person over time.
Ancestry. The cohorts were mostly of European and Asian ancestry, so it is uncertain whether the results apply to people of African ancestry.
Blood only. Blood may not show what immune cells are doing inside tissues.
Estimated inputs. The UK Biobank lacked CD8+ T cell counts, so the team had to estimate them with a model.
Oversimplification. "Good K, bad B" is too neat. Granzyme K granules can also promote inflammation.
What it could change
Topol argues that the TemB+ percentage, or the TemB+ to TemK+ ratio, could become a master indicator of unhealthy immune aging, and a cheap one. He says the researchers are working toward such a test. For people with a poor result he points to lifestyle changes and vaccines, but the article cites no evidence that these lower TemB+ or the risk that comes with it. That remains an open question.
He presents this as a third route toward measuring the immune system, alongside two others he has covered: AI reading of thymus size from low-resolution chest CT scans, and blood-protein "clocks" of immune aging. He says both are moving toward commercial availability. A large, well-validated link to mortality is a real step forward. A test that changes what a doctor should do for a patient has not yet been shown.
Remember
The thymus shrinking matters less than what replaces naive T cells: worn-out granzyme B memory cells or steadier granzyme K ones.
CMV infection and chronic heart or metabolic disease push T cells toward the granzyme B type linked to higher mortality.
The measure did not predict Alzheimer's, Parkinson's, dementia or macular degeneration, so it isn't a general aging gauge.
The 21st Century Cures Act (2016) gave US patients quick access to their records. Many now read lab results, with their normal/abnormal flags, in online portals before a clinician has looked at them.
A typical lab report grid: flags mark values out of range, but not how much each one matters.
A study of more than 600 US clinicians suggests the "normal" and "abnormal" flags on lab reports lead to worse decisions than they need to. Clinicians who were shown how strongly each result shifts the odds of a diagnosis made the right call 93% of the time, compared with 82% for those shown the usual cutoff. Amiran Baduashvili, a University of Colorado physician who teaches diagnostic reasoning, led the work. He notes the cases were simulated.
The case behind it
A patient, Mr. V, arrived with chest tightness, swollen limbs and an enlarged heart on X-ray, all signs of heart failure. His 1, a hormone the heart releases under strain, came back "normal." The resident knew BNP can be falsely normal in overweight patients, so he ignored it and gave a 2 to remove fluid. Mr. V actually had 3: fluid around the heart squeezing it. A diuretic can be deadly in that situation. Surgeons drained the fluid in time.
The clue was in the number. BNP under 100 pg/mL counts as normal, and overweight heart-failure patients usually sit just below that. Values under 50 are rare in heart failure. Mr. V's was 14, which argued strongly against the diagnosis. The label treated 14 the same as 95.
Cutoffs hide direction and strength
Baduashvili gives more examples. A 4 (clot test) between 500 and 1,000 ng/mL is flagged abnormal, but he says it slightly lowers the odds of a clot. 5 cholesterol under 100 mg/dL is "normal," yet above 55 is too high for someone who has already had a heart attack or stroke.
The fix tested
An 6 splits a test's results into ranges. For each range it compares how often people with and without the disease land there. A ratio well below 1 argues against the disease, and near 1 means the result tells you little. In the trial, clinicians at four academic centers read two cases. All saw the exact value, along with either the normal cutoff or the ratios. The ratios helped across experience levels.
Caveats
The decisions were made on vignettes, not real patients. The article doesn't name the tests used or the journal. Health records show only cutoffs today, and good source data for the ratios is hard to find. Baduashvili founded a company that sells a diagnostic-reasoning course, though he says he earns nothing from it. His advice to patients: ask what your specific number means, not just whether it's flagged.
Amiran Baduashvili, who teaches diagnostic reasoning at the University of Colorado Anschutz, argues that the "normal" and "abnormal" flags on lab reports hide the information doctors most need. His team ran a randomized study of more than 600 clinicians to test an alternative. Clinicians shown 1, which say how strongly a particular range of values points toward or away from a diagnosis, made the right management call 93% of the time. Those shown the usual normal-range cutoff got it right 82% of the time. The cases were simulated, not real patients.
A near-fatal misreading
Baduashvili starts with a real case. Mr. V came to an emergency department with chest tightness and shortness of breath. He had swollen hands and legs, and his chest X-ray showed an enlarged heart. Everything suggested heart failure, where a weak pump lets fluid build up in the body.
One result didn't fit. 2 (B-type natriuretic peptide), a hormone the heart releases when it is strained, came back normal. The resident remembered, correctly, that BNP can be falsely normal in people with obesity or excess weight who do have heart failure. So he set the result aside and ordered a 3, a drug that removes fluid. That is standard treatment for heart failure.
Mr. V didn't improve. The next day a supervising physician did a bedside ultrasound and found 4. In this condition, fluid in the sac around the heart squeezes it so hard it can barely fill. A compressed heart needs more blood volume, so the diuretic could have killed him. Surgeons drained the fluid and he survived.
Why "normal" misled
BNP under 100 pg/mL (picograms per milliliter) counts as normal. Overweight patients with heart failure often fall below 100, but usually only just. Values under 50 are uncommon in heart failure, even in very overweight patients. Mr. V's BNP was 14. That number didn't just fail to support heart failure. It argued strongly against it. The label put 14 in the same box as 95, and the clue that should have sent doctors looking for another cause was lost.
The general point is that a single cutoff throws away how far a result sits from the line. Baduashvili gives two more examples:
5, a blood test used to look for clots, is flagged abnormal above 500 ng/mL. He says values between 500 and 1,000 actually slightly lower the odds of a clot. For a low-risk patient, a result in that range is almost as reassuring as a normal one, even though the report flags it.
6 cholesterol under 100 mg/dL is labeled normal. For someone who has already had a heart attack or stroke, anything above 55 is too high.
His analogy: no doctor would accept a blood pressure cuff that only said "elevated," or a heart monitor that only said "low." Yet lab reports are effectively that.
How interval likelihood ratios work
To build them, researchers run a test on many patients, some with the target disease and some without. Instead of drawing one line, they split the results into ranges. For each range they compare how often people with the disease land there versus people without it. That ratio is the likelihood ratio for the range.
A worked example, using made-up numbers: suppose 2% of heart-failure patients have BNP below 50, compared with 40% of patients with other causes of breathlessness. The likelihood ratio for that range is 0.02 ÷ 0.40 = 0.05. Doctors multiply their prior odds by this figure. If a clinician thought the odds of heart failure were 3 to 1 (75%), a ratio of 0.05 brings them to 0.15 to 1, about 13%. Now another diagnosis needs to be found. The same patient with BNP of 95 might get a ratio near 1, which barely moves the estimate. Both values are "normal," but they mean very different things.
The study
The team recruited more than 600 clinicians at four major US academic institutions: attending physicians, residents and 7 such as nurse practitioners. Each read two scenarios like Mr. V's and made a key management decision based on a lab result. Everyone saw the exact number. Participants were randomly assigned to see it alongside either the standard normal cutoff or interval likelihood ratios.
Correct decisions were 93 in 100 with likelihood ratios and 82 in 100 with the cutoff. That is an 11-point absolute gain, or roughly one better decision for every nine or ten patients like Mr. V. Baduashvili reports that the benefit held across experience levels and training backgrounds.
He also points to a legal side. An "abnormal" flag can make a clinician feel obliged to order more tests, such as a scan, even when their judgment says it isn't needed. A displayed likelihood ratio gives a documented, evidence-based reason to hold off, which spares patients the risks of those tests.
Caveats
Simulated, not real. Baduashvili says this himself. Better choices on paper vignettes may not carry over to busy wards, and the next step is testing in real practice across institutions, settings and tests.
Details are thin. The article doesn't name the journal, which tests were used, or how the "correct" answer was defined. It also says nothing about whether the gap was similar in both scenarios.
Data availability. Interval likelihood ratios require good studies for each test in the relevant population. Electronic health records currently show only binary cutoffs, and the author admits that finding high-quality source data is a real obstacle.
Disclosure. Baduashvili founded EBM Clarity, Inc., which owns an online course in diagnostic reasoning. He says he earns no income from it.
What patients can do
Baduashvili's practical advice is to treat a flag as the start of a conversation. Ask your clinician what your specific number means for you, not just whether it fell inside the range.
Remember
Two "normal" results can mean opposite things: BNP of 14 nearly rules out heart failure, while 95 barely changes the odds.
Some mildly "abnormal" flags, like D-dimer of 500–1,000 ng/mL, can actually make the disease slightly less likely.
The 11-point accuracy gain came from vignettes, so it still needs testing with real patients.
The clinic opening at NIH: parents and officials, most not publicly identified, at the vaccine injury clinic sign.1 / 2
The news
On October 5, Health and Human Services (HHS) Secretary Robert F. Kennedy Jr. opened a "vaccine injury clinic" at the National Institutes of Health (NIH). The roundtable that went with the opening was held behind closed doors. HHS waited until October 8 to announce it, after the New York Times and Wall Street Journal had already reported it. The press release lists four actions:
an NIH clinic to study and treat people with health problems after vaccination
a proposed Medicare policy to pay physicians for reporting suspected injuries to 1, the Vaccine Adverse Event Reporting System
a requirement that electronic health records capture vaccine adverse events
an upgraded VAERS, including a mobile version and a QR code on the information sheets patients get with each shot
The critique
David Gorski, a surgeon who edits Science-Based Medicine, says that taken alone, most of these ideas would be fine "under a normal administration." His objection is to who is running them. Several roundtable guests are linked to 2, a group he calls antivaccine. Others are parents whose claims of vaccine injury Gorski disputes. One mother calls her son's paralysis polio caused by a 3 shot, but that vaccine contains no polio component. Gorski says he believes these families' accounts of what happened, but not necessarily their belief that vaccines were the cause.
Paid reporting. VAERS takes reports from anyone, and a report doesn't show cause. Gorski points to California: after it ended nonmedical school exemptions, a cottage industry of doctors selling medical exemption letters appeared. He predicts paid reporting will do the same to VAERS: more reports, no more real injuries, and real signals harder to find.
Records mandate. The 4 already scans health records at large health systems, mainly Kaiser Permanente, for problems that cluster after vaccination. Kennedy hasn't been able to get access to its data. Gorski reads the new records requirement as an attempt to build his own version "by fiat" across every health system. Wider coverage would help detect rare events in principle, he grants, but he doesn't trust this HHS to do it.
What's missing
No director, budget, enrollment rules or oversight structure for the clinic has been announced. HHS hasn't answered press questions about them. Outside scientists such as Vanderbilt's Kathryn Edwards told Science the effort could be worthwhile only if it is done rigorously.
This is commentary, not a study. The facts about the rollout are documented. The claim that the program is a "fishing expedition" for evidence against vaccines is Gorski's prediction about intent, based on Kennedy's record. Things to watch: who is named to lead the clinic, and whether VAERS reports rise while confirmed injury rates do not.
What was announced
On October 5, Health and Human Services (HHS) Secretary Robert F. Kennedy Jr. cut the ribbon on a "vaccine injury clinic" at the National Institutes of Health (NIH) campus in Bethesda, Maryland. Dr. Mehmet Oz, who heads the Centers for Medicare & Medicaid Services (1), attended, as did new CDC director Dr. Erica Schwartz and acting FDA commissioner Kyle Diamantas. The roundtable that went with the opening was held behind closed doors. NIH staff were encouraged to work from home that day. HHS didn't issue its press release or social media posts until October 8, after the New York Times and Wall Street Journal had already reported the event.
The press release lists four actions:
An NIH clinic "dedicated to studying and caring for patients experiencing health problems following vaccination."
Payments to physicians for reporting suspected vaccine injuries to 2, the Vaccine Adverse Event Reporting System, under a proposed CMS policy.
A requirement that electronic health records capture vaccine-related adverse events, set through federal health IT standards from the Office of the National Coordinator for Health Information Technology.
An upgrade to VAERS. On September 28 the CDC awarded a contract to make the system usable on mobile devices and improve its tools for researchers. A VAERS QR code has also been added to the Vaccine Information Statements that patients receive with each shot.
David Gorski, a surgeon and the managing editor of Science-Based Medicine, writes that taken alone, most of these ideas would be reasonable "under a normal administration." His case is about who is running them and how the launch was handled. Much of his piece is opinion, and he says so openly.
Who was in the room
Gorski identified several roundtable guests from photos and posts by allied groups. Ernesto Ramirez Sr. appeared in the 2023 film Shot Dead, which blamed his son's sudden death on the COVID-19 vaccine. Gorski says the autopsy found an anomalous coronary artery, a birth defect that can cause sudden cardiac death in teenagers. Brianne Dressen sits on the board of 3, a group Gorski calls antivaccine. She says the AstraZeneca COVID-19 vaccine trial disabled her, and she is suing the company. Neurologist Suzanne Gazda sits on React19's medical board and received an award from the 4, a group that promoted unproven COVID-19 treatments.
The Defender, the news site of Children's Health Defense (the antivaccine group Kennedy used to lead), quoted two more parents. One says her son's paralysis was "vaccine-induced polio" after a 5 shot (diphtheria, tetanus and pertussis). Gorski points out that DTaP contains no polio vaccine. The only polio vaccine that can rarely cause polio, the oral one, was dropped in the US more than 20 years before the boy was born. The family has also said the CDC hasn't confirmed his diagnosis. The other parent's son died about 14 hours after a flu shot. His death was reported to VAERS and investigated, and no link to the vaccine was found.
Gorski says he believes these families' accounts of what happened. What he disputes is their conclusion that vaccines were the cause. He reads the same split into Schwartz's pledge that the CDC will "listen, investigate, believe them." In antivaccine rhetoric, he argues, "believe the parents" means accepting their view of the cause, not just their account of events.
Why paying for VAERS reports worries him
VAERS is passive: it records whatever is submitted. That makes it uneven. Dramatic events like deaths and seizures are reported reliably, while mild reactions are underreported. Its openness also lets activists flood it. In one well-known case, someone reported that a flu shot had turned him into the Incredible Hulk. Gorski says litigation over vaccines and autism skewed VAERS in the early 2000s, and that activists later used it against COVID-19, HPV and MMR vaccines. A "Harvard study" claiming that only 1% of reactions get reported was cited at the event. Gorski calls that reading a gross exaggeration.
His prediction comes from an analogy. After California's 6 ended nonmedical school vaccine exemptions, a "cottage industry" of doctors sprang up selling medical exemption letters. He expects paid reporting to do the same: more VAERS reports, with no rise in true adverse events, so real signals get harder to find. He does sympathize with the point about unpaid paperwork. He also notes that a new diagnostic code for vaccine injury, pushed by React19, was proposed earlier this year even though codes for vaccine reactions already exist.
The records mandate and the 7
The Vaccine Safety Datalink (VSD) is active surveillance. It automatically scans patient records at participating health systems for events that appear more often soon after vaccination. The data stay on the health systems' own servers, and Kaiser Permanente guards them closely. Gorski says Kennedy has wanted this data for a long time. He hired David Geier, whose VSD access was revoked in 2003 over privacy concerns, yet so far appears unable to get any data after 2001. Gorski reads the new records requirement as an attempt to "reconstitute his own VSD by fiat" across every health system.
He grants that in principle this is a good idea. The VSD is sometimes too small to detect rare events, and wider coverage would help. He objects to who would be running it.
What's missing
No director, budget, enrollment criteria or oversight structure for the clinic has been announced. Gorski says he can't remember another clinic launch that didn't name its leader. News outlets report that HHS hasn't answered questions about these details. There is also no published way for doctors to refer patients.
Outside scientists were cautiously positive. Vanderbilt pediatrician Kathryn Edwards told Science the effort was "laudable" but would be "antivaccine propaganda" if not done scientifically. Daniel Salmon of Johns Hopkins told the New York Times more safety research would help. The Defender left out the rest of his quote: "It needs to be rigorous science."
How to weigh this
This is a commentary from a long-standing critic of Kennedy, not a study. Its factual core is well documented: the delayed and low-key rollout, the guest list, the four policy items, and the missing leadership and budget. Its central claim, that the program is a "fishing expedition" meant to produce evidence against vaccines, is a prediction about intent. Gorski supports it with Kennedy's track record, including a recent speech linking vaccines to diabetes, SIDS, asthma and autoimmune disease. Gorski also restates the mainstream view that serious vaccine injuries are real but rare, and far smaller than the harms vaccines prevent. He says he would accept a rigorous finding against a vaccine. The real tests come later: who is named to lead the clinic, which contractor rebuilds VAERS, and whether VAERS report counts rise while confirmed adverse event rates do not.
Remember
Paying doctors for VAERS reports could raise report counts without any rise in real injuries, so the two must be tracked separately.
The new health-record requirement could widen vaccine safety monitoring, which helps or hurts depending on how the data are analysed.
Watch who is named to lead the clinic and which contractor rebuilds VAERS. Gorski calls these the real tells.
A new mouse study offers one explanation for why senescent cells build up in old tissue. The blog Fight Aging! reports on it. The paper blames the age-related decline of 1, a pathway that picks out individual proteins and sends them to the 2, the cell's recycling compartment, to be broken down. When the researchers boosted CMA with a drug, old mice had fewer senescent cells. Mice with lung scarring had milder disease.
The problem
Cells become senescent after damage or stress, or after reaching the 3, the cap on how many times a cell can divide. They stop dividing and release inflammatory signals called the 4 (senescence-associated secretory phenotype). In youth, immune cells clear them quickly. In old age, clearance fails and the cells linger. Their signals harm tissue even when there are few of them.
Researchers haven't settled whether the immune system is to blame or whether old senescent cells become harder to clear. This paper suggests the two problems are connected.
How CMA works
Proteins marked for CMA carry a short tag, the KFERQ-like motif. A chaperone protein, 5, picks up tagged proteins and brings them to 6, a receptor on the lysosome. LAMP2A molecules gather into a channel that feeds the protein inside to be destroyed. CMA weakens with age in most tissues.
What they found
Cells without CMA still become senescent, but they look like aged senescent cells in their proteins, metabolism and secretions.
Their secretions push nearby cells toward senescence. They also switch off CMA in 7, the immune cells that engulf senescent cells. Macrophages without CMA are worse at engulfing them.
In mice, blocking CMA only in macrophages increased senescent cell build-up and slowed clearance during wound healing.
A CMA-activating drug lowered senescent cell burden in aged mice. It also eased disease in a mouse model of pulmonary fibrosis, a scarring lung disease.
Caveats
This is mouse work only. The post quotes just the abstract, so it gives no effect sizes and doesn't name the drug. The approach differs from 8, drugs that kill senescent cells: the aim is to restore the body's own clearance. Reason notes that CMA involves many steps, which gives drug developers several possible targets. Whether any of it carries over to humans is untested.
The finding
A newly published, open-access paper reports a mouse study of why senescent cells pile up in old tissue. Reason covers it on the longevity blog Fight Aging!. The proposed culprit is 1, a selective protein-recycling pathway that weakens with age. According to the paper, falling CMA changes senescent cells themselves. It also disables CMA in 2, the immune cells that would normally eat them. When the researchers activated CMA with a drug in old mice, senescent cell numbers fell. In a mouse model of lung scarring, disease was less severe.
The problem: why do senescent cells linger?
A cell becomes senescent after stress, damage or injury, or most often after hitting the 3, the cap on how many times a cell can divide. A senescent cell stops replicating, grows larger and releases a strong mix of pro-growth and pro-inflammatory signals. This mix is called the senescence-associated secretory phenotype, or 4.
Senescent cells form throughout life and are useful in the short term. They help coordinate wound repair and draw immune attention to cells that might become cancerous. In young bodies the immune system clears them efficiently. In old bodies it does not, and a lasting population builds up. Reason argues that even a small one, relative to all the cells in a tissue, disrupts tissue structure and function through constant signalling.
The open question is where the failure lies. Does the aging immune system get worse at clearance? Or do aged senescent cells change in ways that make them harder to clear? Reason notes there is some evidence for both. This paper suggests the two are linked.
How CMA works
In CMA, the cell removes individual proteins one at a time. The steps are:
Tagging. Proteins eligible for CMA carry a short amino-acid sequence called a KFERQ-like motif.
Pickup. A chaperone protein called 5 (heat shock cognate protein of 71 kDa) recognises the motif and binds the protein.
As a rough analogy, the motif works like a shipping label, HSC70 like a courier, and LAMP2A like a loading dock that has to be assembled for each delivery. Earlier work, summarised in the paper, found that CMA declines with age in most tissues. That decline has been linked to neurodegeneration, the metabolic syndrome of aging and atherosclerosis.
What the researchers showed
The abstract describes a chain of effects:
CMA loss changes senescent cells. Cells lacking CMA still become senescent, but their proteins, metabolism and secretions look like those of aged senescent cells, not young ones.
Their SASP spreads and suppresses clearance. The secretions of these cells pushed neighbouring cells further toward senescence. They also inhibited CMA inside macrophages.
Macrophages without CMA eat less. With CMA suppressed, macrophages were worse at engulfing senescent cells.
Removing macrophage CMA in mice had visible effects. Blocking CMA only in macrophages increased senescent cell build-up. It also slowed the clearance of senescent cells during wound healing.
Activating CMA helped. A drug that activates CMA reduced the burden of senescent cells in aged mice. It also reduced disease severity in a mouse model of pulmonary fibrosis, a disease in which lung tissue scars and stiffens.
The authors conclude that CMA decline drives senescent cells to persist and that raising CMA is a promising way to resolve senescence in old organisms.
Caveats
Mice only. None of this has been tested in people.
No numbers here. The post quotes only the abstract. It gives no effect sizes, sample sizes or measures of how much senescent burden fell, and it does not name the CMA-activating drug.
Partial mechanism. The study shows that the SASP of CMA-deficient cells suppresses macrophage CMA. The material here does not identify which secreted factors do it.
Unclear share of the problem. It is not clear how much of the age-related failure of clearance this pathway explains, compared with other forms of immune aging.
Why it's interesting
Most anti-senescence drug work so far has focused on 8, drugs that kill senescent cells directly. This study points to another approach: restore the body's own clearance by repairing a maintenance pathway in both the senescent cells and the immune cells. Reason adds that CMA involves many steps, so drug developers could target several of them. Whether that works in humans, and at what doses, remains unknown.
Remember
Failing CMA in aged senescent cells makes their secretions switch off CMA in nearby macrophages, so clearance fails.
Removing CMA only in macrophages was enough to make senescent cells build up in mice.
A CMA-activating drug reduced senescent cells and lung fibrosis in mice. There are no human data, and the post reports no effect sizes.
Setting: 29 German sites, 360 patients, randomized 2:1 to repair or medical therapy. It was 2, so patients and doctors knew who got the procedure, and there was no sham procedure.
Patients: average age about 80, and sicker than in earlier trials. 55% had been hospitalized for heart failure in the previous year, 82% had 3 (heart and kidney failure worsening each other), and 92–98% had atrial fibrillation. Control-arm one-year mortality was about 17.5%, compared with 5–8% in prior trials. Several devices were allowed, but 98% of patients got TEER.
First primary endpoint: a hierarchical composite scored by win ratio. Every repair patient is compared with every control patient. Within each pair, the patient who lived longer "wins". If that is tied, the one with fewer heart-failure hospitalizations wins. If that is also tied, the one with more quality-of-life improvement at one year wins. The win ratio is total repair wins divided by total control wins.
Second primary endpoint: tested only if the first succeeded. It was death from any cause or heart-failure hospitalization through three years.
52.4% vs 21.0%
5
Heart-failure hospitalization alone: 19.4% vs 43.9% (HR 0.35). TRILUMINATE showed no such effect.
Deaths at one year: 13.1% vs 16.3% (HR 0.74, CI 0.42–1.29, not significant). At three years: 19.0% vs 27.6% (HR 0.62, CI 0.40–0.96).
Major adverse events within 30 days: 14 patients (6%) in the repair arm.
Docking. HSC70 delivers the protein to 6 (lysosome-associated membrane protein 2A), a receptor on the surface of the 7.
Entry. LAMP2A molecules cluster into a channel-like complex, and the protein is threaded into the lysosome to be broken down.
