FDA approves first factor XIa blood thinner, cutting repeat strokes 26%
Nov 10, 2026
Listen to this post
read by AI voices · all episodesTHE SHORT READ
The Food and Drug Administration on Monday approved asundexian, the first of a new class of anticoagulants, for adults with a recent noncardioembolic ischemic stroke or high-risk transient ischemic attack, taken alongside antiplatelet therapy.
In a 12,327-patient trial, the drug cut ischemic stroke 26% against placebo while overall major bleeding was similar, the agency said, though bleeding rose in patients older than 75 and in the first 90 days. Bayer set a list price of $748.50 for 30 days, BioPharma Dive reported.
An NEJM AI analysis of Medicare billing found family and internal medicine have the least physical protection from AI substitution. A study in Annals of Internal Medicine tied first-trimester doxycycline to more pregnancy losses, and a Brazilian trial in NEJM found a systolic target under 120 mm Hg did not reduce cardiovascular events in type 2 diabetes.
THE DETAILS
- FYIRegulation Nov 9, 2026The Food and Drug Administration approved Bayer's Sunharbo (asundexian) 50 mg once daily on Nov. 9 for adults with a recent mild-to-moderate noncardioembolic ischemic stroke or a high-risk transient ischemic attack, taken with antiplatelet therapy. In the double-blind OCEANIC-STROKE trial of 12,327 patients randomized within 72 hours of symptom onset and treated a median 14.2 months, it reduced ischemic stroke 26% versus placebo (hazard ratio 0.74, 95% CI 0.65 to 0.84, per a published review); overall major bleeding was similar, but major bleeding rose in patients older than 75 and during the first 90 days, and gastrointestinal bleeding was more common, the agency said. The label calls for holding the drug at least 24 hours before elective procedures and after thrombolysis, and it is not approved for atrial fibrillation, where asundexian proved inferior to apixaban in an earlier trial. Bayer set a list price of $748.50 for 30 days and plans to reach stroke neurologists first and primary care physicians later, BioPharma Dive reported.
- FYIWorkforce Nov 4, 2026Only about one fifth of Medicare Part B payments go to services other than office visits that carry no physical barrier to AI substitution, according to an analysis of 2023 claims, nearly $45 billion paid to 589,648 physicians in 56 specialties, published Nov. 4 in NEJM AI. The Stanford and Berkeley authors classified every billed service except evaluation and management visits by whether it requires physical contact, and they describe that physical share as a ceiling on how much of a specialty's work AI alone could replace. Procedural fields such as dermatology and orthopedic surgery were largely shielded, and protection rose with the length of postgraduate training. Family and internal medicine, where physicians concentrate, had low protection by that measure, and the authors said whether AI substitutes for or complements their work will depend as much on how it is designed, deployed and paid for as on the technology.
- FYIEvidence Nov 10, 2026Pregnancies exposed to oral doxycycline in the first trimester for a respiratory tract infection ended in loss more often than those exposed to amoxicillin (adjusted risk ratio 1.54 in commercial claims and 1.40 in Medicaid) or a macrolide (1.39 and 1.36), according to a CDC-funded active-comparator cohort study published Nov. 10 in Annals of Internal Medicine. Among 6,636 pregnancies with acne, doxycycline versus topical clindamycin carried higher risk in commercial claims (1.49) but not in Medicaid (1.03, 95% CI 0.80 to 1.33). Where the risk ratios were significant, absolute risk differences ranged from 7 to 11 percentage points. The authors noted possible residual confounding and misclassification, did not study malformations, and wrote that clinicians may use the findings to weigh risks and benefits when prescribing doxycycline to reproductive-aged women.Sources: Annals of Internal Medicine, PubMed
- FYIEvidence Nov 2, 2026Inference-time activation steering, which alters a language model's internal activations without changing its weights or system prompt, shifted an open-weights model's depression advice between antidepressants and self-care, Dr. Roy Perlis of Massachusetts General Hospital reported Nov. 2 in JAMA Network Open. Across 372 responses from DeepSeek version 4 Flash to 12 synthetic scenarios, the balance of medication versus self-care content fell 0.32 points per unit of steering on a scale from -3 to 3. The shift was largest when the simulated patient stated no treatment preference, it did not override explicit requests for or against medication, and 87% of responses still advised seeing a clinician. Perlis wrote that the change would be invisible to a patient or clinician and that the results suggest a need for independent auditing if models inform clinical decisions; how larger commercial models respond is unknown.Sources: JAMA Network Open, PubMed
- FYIEvidence Nov 7, 2026Targeting systolic pressure below 120 mm Hg instead of below 140 mm Hg did not reduce major cardiovascular events in high-risk adults with type 2 diabetes in OPTIMAL-DIABETES, an open-label randomized trial in Brazil published Nov. 7 in NEJM. Mean systolic pressure during follow-up was 121.8 versus 135.4 mm Hg. Over a median 4.0 years, cardiovascular death, nonfatal myocardial infarction, nonfatal stroke or hospitalization for unstable angina or heart failure occurred in 6.0% of the intensive group and 6.7% of the standard group (hazard ratio 0.90, 95% CI 0.77 to 1.06). Serious adverse events were similar at 14.9% and 14.6%; 62.0% of the 9,476 patients were women and the mean age was 64.7.
- FYIEvidence Nov 5, 2026Across 2,080,765 adult encounters at 15 Danish hospitals from 2018 to 2023, a gradient-boosting model built on vital signs, demographics, laboratory values and medical-history text gained at most 1.9 correct 24-hour mortality identifications per 10,000 patients over the National Early Warning Score, according to a retrospective study published Nov. 5 in The Lancet Digital Health. In patients whose first score was 7 or higher, the model gained up to 78.1 per 10,000, while a simplified score without blood pressure and temperature lost up to 60.2 per 10,000; for scores of 0 to 4 the differences were under 1 per 10,000. The authors used propensity weighting to account for treatment given after the first score and called for prospective trials of monitoring stratified by the first assessment.Sources: The Lancet Digital Health, PubMed