The IV-versus-IO question answered as a null in 7,561 patients, peripheral perfusion-guided resuscitation at 97.2% Bayesian probability, barriers to dispatcher CPR, and minimally invasive surgery for mitral endocarditis

The trailing-window rule paid for itself in one day. Yesterday’s brief established that Europe PMC had indexed zero records for 12 September and 1,035 for 11 September, and wrote into the source notes that every run must re-sweep a 2–3 day window rather than only the new date. This morning:

Date	Yesterday	Today
11 September	1,035 records	3,805
12 September	0	1,147
13 September	—	48

All four items below are dated 11 September and none was visible yesterday. Without the trailing window they would have been lost permanently.

Three of the four bear directly on the ERC 2025 resuscitation guidelines backfilled yesterday, which is a coincidence of timing worth taking advantage of.

1. Intravenous versus intraosseous access in cardiac arrest — 7,561 patients, and it is a null

Journal · Published: Resuscitation, 11 September 2026 — two days old Holmberg, Couper, Andersen, Lall, Granfeldt, Ji, Perkins, Vallentin

Systematic review and individual participant data meta-analysis — PRISMA-IPD, searched to 18 March 2026 — of randomised trials comparing intraosseous-first with intravenous-first vascular access during cardiac arrest. Logistic regression adjusted for prognostic factors, prespecified Bayesian analyses, GRADE certainty.

Two randomised trials, individual participant data on 7,561 patients with out-of-hospital cardiac arrest.

• 30-day survival: OR 1.01 (95% CI 0.81–1.25) — no difference, moderate certainty
• Favourable neurological outcome at 30 days or discharge: OR 1.08 (0.83–1.41) — no difference, low certainty
• Sustained ROSC: OR 0.90 (0.81–1.01) — point estimate favours intravenous, not statistically significant, low certainty
• Bayesian analyses found a high probability that any difference between the two strategies was small

Interpretation. This is the definitive version of a question this archive touched yesterday, and it sharpens it considerably.

The ERC 2025 guidelines backfilled yesterday identify intravenous access as superior to other routes of drug administration — one of the headline changes, reversing a decade of drift toward “IO first because it’s faster.” I noted then that the trials showed better ROSC or drug-delivery outcomes without a clear survival benefit. This IPD meta-analysis quantifies exactly how thin that basis is.

Thirty-day survival is flat — OR 1.01, at moderate certainty, in 7,561 patients. That is about as clean a null as resuscitation research produces. Neurological outcome is flat too. The entire signal favouring IV is a non-significant ROSC difference whose confidence interval touches 1.01, at low certainty. And the Bayesian analysis says the honest thing: whatever difference exists is small.

So how should the ERC recommendation be read? Not as wrong — a real ROSC advantage with no survival translation is a legitimate basis for preferring a route, especially when neither route carries extra risk and IV is universally available. But it is a recommendation about process on weak grounds, not a survival intervention. The practical consequence is unchanged from yesterday and now better supported: prefer IV where you can get it quickly, and do not delay drugs to obtain it — because the survival data give you no reason to trade time for route.

Two limits worth naming. Only two trials contributed IPD — PARAMEDIC-3 and IVIO — so “individual participant data meta-analysis of 7,561 patients” is really two large trials analysed together, which is a strength for consistency and a limit for generalisability. And this is out-of-hospital arrest only. For in-hospital arrest, where IV access usually already exists and the question rarely arises, none of this applies directly.

Read the meta-analysis · PMID 42727699

2. Peripheral perfusion-guided resuscitation in septic shock — 97.2% Bayesian probability of benefit, and a frequentist P of 0.06

Journal · Published: Critical Care Medicine, 11 September 2026 — two days old Mendes, Silva Correia, Moreira Salles, Kattan, Hernández, Melo · PROSPERO CRD420251229564

Systematic review and prespecified dual-framework meta-analysis — a primary frequentist random-effects model and a Bayesian hierarchical model with alternative priors — of RCTs comparing peripheral perfusion-guided resuscitation with standard care or lactate-guided resuscitation. Four databases to November 2025. Cochrane RoB 2.0, GRADE. Note the author list: Kattan and Hernández are the ANDROMEDA-SHOCK investigators, which makes this a synthesis by the people whose trial dominates it.

Seven RCTs, n = 2,408. All enrolled patients with septic shock.

• Frequentist: pooled RR for 28-day mortality 0.87 (95% CI 0.76–1.01), P = 0.06, moderate-certainty evidence — the interval crosses no effect
• No heterogeneity at all: I² = 0%
• The two largest and lowest-risk-of-bias trials were individually non-significant
• Bayesian: 97.2% posterior probability that RR < 1, broadly consistent across alternative priors
• Authors’ conclusion: directionally concordant, “although the magnitude of any effect remains uncertain,” and adequately powered trials are needed

Interpretation. This is the most interesting methodological object of the day, and the authors have handled it well.

A pooled RR of 0.87 with I² = 0% and P = 0.06 is the classic shape of a real, modest effect in an underpowered evidence base. Zero heterogeneity across seven trials means they are all telling the same story; 2,408 patients is simply not enough to resolve a 13% relative reduction. The frequentist framework calls that “not significant”; the Bayesian framework calls it “97.2% probable.” Both are correct, and the disagreement is about the question, not the data.

Two things stop this being a simple “so it works” reading, and to their credit the authors flag both. The two largest and least-biased trials were individually null — so the pooled estimate is being carried in part by smaller, higher-risk trials, which is the standard way meta-analytic optimism gets manufactured. And the comparator is mixed: standard care or lactate-guided resuscitation. Beating lactate-guided care and beating usual care are different claims, and pooling them obscures which one the 0.87 refers to.

Where this sits in the archive. This is the fifth item in a month converging on the same idea, and it is now the most direct statement of it. The renal perfusion phenotype cohort on 8 September found a semi-quantitative perfusion score discriminated persistent AKI while renal resistive index trajectories did not. The Hypotension Prediction Index analysis showed a pressure-targeting technology’s benefit was really treatment intensity. The ETCO₂ cohort found a perfusion-adjacent variable carrying prognostic information independent of blood pressure. Yesterday’s ADQI/ELSO ECMO consensus named venous congestion among AKI drivers. Perfusion, not pressure has gone from a slogan to a consistent thread, and this meta-analysis is the closest thing to a mortality argument for it — at 97.2% posterior probability and P = 0.06.

Practically: capillary refill time is free. On this evidence it is at least as good a resuscitation target as lactate, plausibly better, and the downside is nil. That is a reasonable thing to act on before the adequately powered trial arrives.

Read the meta-analysis · PMID 42725818

3. Barriers to dispatcher-assisted CPR — 43.7% of recognised arrests, and survival halved

Journal · Published: Resuscitation, 11 September 2026 — two days old Boberg, Jonsson, Riva, Cronhjort, Ringh, Nord, Claesson, Robertsson, Merigo, Magliocca, Ristagno, Nordberg, Forsberg, Hollenberg (Karolinska, with Milan)

Retrospective cohort of transcribed 112 emergency calls from five out-of-hospital cardiac arrest studies, linked to the Swedish Registry for Cardiopulmonary Resuscitation. Crucially, calls were included only where the dispatcher had already recognised cardiac arrest — so this measures what goes wrong after recognition. Barriers counted: caller not with the patient, caller left or hung up the phone, language barriers, overly distraught caller, inability to move the patient, CPR refusal.

3,263 confirmed out-of-hospital cardiac arrests.

• 43.7% had at least one barrier
• Most common barrier: inability to move the patient — 55.8% of barrier-present calls
• Cumulative bystander CPR initiation: 85% in barrier-free calls; below 50% with ≥3 barriers
• ROSC: adjusted RR 0.89 (95% CrI 0.81–0.98)
• 30-day survival: adjusted RR 0.57 (95% CrI 0.46–0.71)

Interpretation. The ERC 2025 guidelines expand the dispatch centre’s role in helping the caller recognise arrest and ventilation failure. This paper is about the step after that, and it is where nearly half of everything fails.

A 43% barrier rate in calls where the dispatcher already got the diagnosis right is the headline, and it reframes what dispatcher-assisted CPR training should target. Recognition is the part that gets studied and taught; this says recognition is not the bottleneck in almost half of cases.

The most common barrier is physical, not communicative: inability to move the patient, in 56% of barrier-present calls. That is a person wedged between a bed and a wall, in a bathroom, in a car seat — and no amount of dispatcher scripting solves it. It is the single most actionable finding here precisely because it is so concrete: dispatchers can be trained to coach moving the patient, and to recognise early when the caller cannot and needs different instructions.

The survival number is large — adjusted RR 0.57, a 43% relative reduction. Treat it with the caution any observational association of that size deserves: a call with three barriers is plausibly also a call with an unwitnessed arrest, an elderly patient alone, a delayed recognition upstream. The adjustment cannot fully separate the barriers from the circumstances that produce them. But the dose-response across barrier count — 85% down to under 50% bystander CPR — is the kind of gradient that is hard to explain by confounding alone.

This is the least directly clinical item of the day and arguably the highest-yield: it concerns the first 3–5 minutes that ERC 2025 explicitly strengthened, and it is a problem fixable by changing what dispatchers say.

Read the paper · PMID 42727702

4. Minimally invasive surgery versus sternotomy for mitral valve infective endocarditis

Journal · Published: Perfusion, 11 September 2026 — two days old Motawea, Blackledge, Al-Sheikh, El Diasty, Elgudin, Sabik, Abu-Omar

PRISMA-compliant meta-analysis of randomised trials or cohort studies comparing minimally invasive surgery with full sternotomy for mitral valve infective endocarditis.

Four cohort studies, 466 patients — 137 minimally invasive, 329 full sternotomy. No randomised trials were found.

No significant difference in:

• In-hospital/30-day mortality — RR 0.71 (0.31–1.60)
• Stroke — RR 0.63 (0.16–2.46); atrial fibrillation, AKI/renal failure, multi-organ failure, readmission
• Reoperation, permanent pacemaker, survival to 2.5 years, endocarditis recurrence to 3.5 years

Significantly favouring minimally invasive:

• Postoperative sepsis — RR 0.17 (0.03–0.84), P = 0.03
• Blood transfusion — RR 0.63 (0.48–0.84), P = 0.001
• ICU length of stay — −1.57 days (−2.16 to −0.97), P < 0.00001

Interpretation. Read the sepsis result with real scepticism and the rest as reasonable.

RR 0.17 for postoperative sepsis, with a confidence interval from 0.03 to 0.84, is not a usable estimate. An interval spanning a 33-fold range means very few events; in 466 patients across four cohorts, it may rest on single figures. The point estimate implies an 83% reduction in sepsis from a smaller incision in patients who already have infective endocarditis, which is not biologically impossible but is a very large claim from very little data. I would not quote this number.

The transfusion and ICU stay findings are more credible and consistent with the MiECC meta-analysis sent yesterday from the same lead author and group: less bleeding, fewer transfusions, faster early recovery. That is the reproducible signature of less-invasive cardiac surgery, and it showed up in both analyses.

The decisive limitation is selection, and it is severe here. These are four cohort studies, no randomised trials, and 137 minimally invasive versus 329 sternotomy patients. In mitral endocarditis, the choice of approach is driven by the extent of infection — abscess, root involvement, multivalve disease, haemodynamic instability all push toward sternotomy. The minimally invasive group is by construction the group with less destructive disease. That confounding explains less sepsis, less transfusion and shorter ICU stay without any effect of the incision at all. The authors’ conclusion — “surgical approach should be tailored to the patient’s clinical risk profile, the extent of endocarditis, and the team’s experience” — is correct and is also a description of the confounding.

It does pair usefully with the AHA infective endocarditis scientific statement sent on 9 September, which emphasised the endocarditis team and noted percutaneous aspiration of right-sided vegetations as an emerging option. Both point the same way: the management of endocarditis is diversifying, and the decision about how to operate is becoming as multidisciplinary as the decision about whether.

Read the meta-analysis · PMID 42723536

---

Notes for the next run

• The trailing-window rule works — keep it, and extend it. 11 September grew from 1,035 to 3,805 indexed records in 24 hours. Sweep the last three days every run, and expect the most recent two to be incomplete. 13 September currently has 48 records and will fill in over the next two days.
• Seen on 11 September, worth a slot on a quiet day — all in tracked journals, none yet read beyond the title: Resuscitation on in-hospital cardiac arrest in Impella-supported cardiogenic shock (J-PVAD registry), on first post-resuscitation PaCO₂ and neurologic outcome in children, on hypotension burden from intermittently documented blood pressures after paediatric arrest, and on prognostication after paediatric arrest with reduced EEG electrodes; Critical Care Medicine on a systematic evidence map of biomarker use for AKI and on comprehensive driving assessment before driving resumption after critical illness (two papers); ICVTS on six-minute walk distance and long-term LVAD outcomes in the ELEVATE Registry — which pairs with the DESTINE 2.0 driveline document sent 10 September; Shock on phenotype-guided polymyxin B haemoadsorption in abdominal septic shock; and Perfusion on del Nido cardioplegia in paediatric cardiac surgery, a qualitative sub-study of the DESTINY trial.
• No abstract deposited: the BJA item on interpreting postoperative hypotension (11 September, DOI 10.1016/j.bja.2026.07.055) is a correspondence comment, not the original study — correctly skipped.
• Also seen, out of scope: two AJRCCM papers on fibrotic ILD immunosuppression and Pseudomonas abundotypes in bronchiectasis, a BJA population study on inequality in healthy lifespan after surgery, five Anesthesiology letters in an “Era of Sugammadex” correspondence exchange, an Annals of Thoracic Surgery PET-based recurrence prediction study in stage IA lung adenocarcinoma, and a commentary on sternal wound infection bundles.