Three studies came out this month. They all challenge what we thought we knew…

Plus, three NICE updates this week that are relevant for primary care:

COBRA Trial  |  N Engl J Med

For over a decade, the question of whether to use apixaban or rivaroxaban for treating acute venous thromboembolism (VTE) has remained. Both are considered efficacious.

The COBRRA trial, published in NEJM in March, is the first randomised comparison of the two most prescribed direct oral anticoagulants (DOACs) for acute VTE. Castellucci and his team enrolled 2,760 patients with symptomatic pulmonary embolism (PE) or proximal deep vein thrombosis (DVT). This study took place at 32 sites in Canada, Australia, and Ireland.

Patients were randomised in a 1:1 ratio to standard dosing regimens:

The primary outcome was clinically relevant bleeding over three months, and the results were striking. Bleeding occured in 3.3% of the apixaban group, compared to 7.1% in the rivaroxaban group. This shows a 54% relative risk reduction (RR 0.46; 95% confidence interval [CI] 0.33 to 0.65; P<0.001). Major bleeding was 0.4% versus 2.4%. There was no significant difference in recurrent VTE between the two groups.

The accompanying NEJM editorial puts it plainly: for many patients with acute VTE, the choice of anticoagulant is no longer a toss-up.

The likely explanation is dosing. Rivaroxaban’s loading phase is 15 mg twice daily for 21 days, compared with apixaban’s 10 mg twice daily for just 7 days. That longer, higher-dose window appears to drive most of the bleeding divergence.

Important caveats. This was an open-label trial with blinded endpoint adjudication. It followed patients for only three months, and the population was mostly white. It is a single study, and we should be cautious about treating it as the final word. This is the first randomised evidence directly comparing these two drugs for VTE. The signal is clear and clinically important.

The COBRRA-AF trial, comparing the same two drugs in atrial fibrillation (AF), where there is no loading dose, is underway. This will help show if the safety gap continues in another clinical setting or if the dosing strategy is the main factor.

The PRUDENCE trial tested point-of-care diagnostics across 2,639 patients in 13 countries. It didn’t reduce antibiotic prescribing. The reason why matters more than the result.

Point-of-care testing has been highlighted for years as a vital tool to cut down on unnecessary antibiotic prescriptions. The logic is simple: if clinicians get a quick C-reactive protein (CRP) result, a group A streptococcus test, or an influenza result during the visit, they’ll prescribe fewer antibiotics for cases that are actually viral.

The PRUDENCE trial, published in The Lancet Primary Care on 17 March, tested that assumption across 13 European countries. The results are worth sitting with.

This was a practical randomised controlled trial (RCT) involving 2,639 patients aged one year and older. They had a respiratory tract infection (RTI) and were seen in primary or long-term care. Clinicians were already thinking about or had planned to prescribe an antibiotic. Patients were randomised into two groups. One group received a tailored point-of-care testing strategy. This included CRP for cough, group A streptococcus for sore throat, and influenza A/B during flu season. The other group received usual care only.

The primary outcome was antibiotic prescribing. There was no difference between the two groups. Point-of-care testing did not reduce it. Patient outcomes were equivalent too. No harm, but no benefit in terms of prescribing behaviour.

The companion qualitative evaluation, published alongside the main trial, is arguably the more revealing paper. Interviews with clinicians and patients across six countries found that test results were frequently overridden. When a CRP came back low but the clinician felt the patient looked unwell, or when the patient clearly expected antibiotics, the test lost. Clinical intuition and social dynamics won.

This is just one trial, so it’s too early to say that point-of-care testing has no role in stewardship. Previous studies, particularly those embedding CRP testing within structured communication training (such as the GRACE-INTRO and IMPAC3T trials), have shown reductions in prescribing. PRUDENCE adds a key point: testing done alone, without a larger framework, might not change behaviour. The technology is not the bottleneck. The bottleneck occurs when the result goes against a clinician’s instinct. It also depends on whether the consultation structure allows for action.

In UK primary care, CRP testing is becoming more common and may even be incentivised. This serves as a helpful reality check. Around 90% of antibiotics are prescribed in primary care, and most are for respiratory tract infections. To tackle antimicrobial resistance (AMR), the PRUDENCE data shows that investing in conversation may be just as important as investing in testing.

Ozempic for Alzheimer’s? The EVOKE trials are in, and the answer is no.

If you’ve had patients ask whether their weight-loss injection might also protect them from dementia, you’re not alone. Recent observational data has linked GLP-1 RAs to a lower risk of Alzheimer’s. This information has been in mainstream media for over a year, and some findings seemed quite promising. A large retrospective study found that semaglutide cut Alzheimer’s risk by 67% compared to insulin. It also lowered the risk by 41% compared to other GLP-1 RAs. Pre-clinical models showed neuroprotective effects through anti-inflammatory and vascular pathways.

Novo Nordisk took the hypothesis seriously and ran two large Phase 3 trials to find out. The results were first reported in November 2025. They were fully presented at the AD/PD 2026 conference in Copenhagen on 19 March. These findings are definitive.

EVOKE and EVOKE+ enrolled 3,808 adults aged 55 to 85. These participants had mild cognitive impairment (MCI) or mild Alzheimer’s disease dementia, and they tested positive for amyloid. This took place across nearly 40 countries. Participants were randomised to oral semaglutide 14 mg daily or placebo for 104 weeks, on top of standard care (including approved Alzheimer’s medications).

On the primary endpoint, change in Clinical Dementia Rating Sum of Boxes (CDR-SB) at two years, there was no difference between semaglutide and placebo. Not a small effect that missed significance. No drug-placebo difference at all. The same was true for every secondary endpoint: activities of daily living, Mini-Mental State Examination (MMSE), and Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog). A pooled analysis found no difference in progression from MCI to mild Alzheimer’s. Subgroup analyses by sex, age, body mass index (BMI), and diabetes status showed no signal of benefit in any group.

There was one positive finding. Semaglutide lowered peripheral inflammatory markers significantly. C-reactive protein (CRP) fell by about 30%. It also caused modest reductions of around 10% in cerebrospinal fluid (CSF) biomarkers linked to Alzheimer’s. The investigators and independent commentators agree: these biomarker changes were too small to lead to clinical benefits, and they didn't.

The principal investigator, Jeffrey Cummings, has been candid about what this means. Semaglutide does not substantially enter the brain. It reduced peripheral inflammation effectively, but that was not enough to alter the course of a central nervous system disease. The finding doesn’t rule out brain-penetrant GLP-1 RAs or anti-inflammatory drugs. But, it does eliminate semaglutide for this use. Novo Nordisk has terminated the programme and discontinued the planned one-year extension of both trials.

These results have not yet been published in a peer-reviewed journal. Full publication is expected but at the time of writing, the data is from conference presentations only. The trial design paper is available in Alzheimer’s Research & Therapy.

For primary care, the practical takeaway is straightforward. GLP-1 RAs remain important medicines for type 2 diabetes and obesity. But if a patient asks whether their semaglutide might help with cognitive decline or dementia risk, the answer based on the best available evidence is that it does not. The observational associations were real, but they did not survive the scrutiny of a well-designed randomised trial.