I am at the Critical Care Reviews 2026 meeting this week. It’s an amazing meeting and I am always hugely impressed by the quality of the presentations, the dedicated time provided for editorial commentary, the linked publication programme, and the actively encouraged audience participation. Hats off to the MacSweeneys as usual. Get to the livestream if you can, but even if not I would strongly encourage a look at the website, a review of the studies presented and a recap on the major findings.
Of particular note to those of us working in EM and ICM was the presentation today of the ARISE FLUIDS trial, looking at the use of early vasopressors with restricted intravenous fluids for sepsis. This is an issue that remains controversial in the UK and more data on effectiveness and safety are certainly welcome. The presentation was linked to a concurrent publication in the NEJM and the paper/supplementary appendices are all now available to review, albeit behind a paywall. The abstract is below and like always, we would strongly suggest you access and read the full paper yourselves. But in the meantime, here are some early thoughts from St Emlyns. We have also asked Alasdair Corfield, Chief Investigator for the currently running EVIS trial in the UK, for his thoughts as an expert in the field. The below is our combined hot take on the study
Background
The optimal approach to early resuscitation in septic shock is unknown. Equipoise exists between the use of larger volumes of intravenous fluids to restore perfusion and the use of early vasopressor therapy along with smaller volumes of fluids to minimise potential harm from excess fluid.
Abstract
Methods: We randomly assigned adult patients who presented to the emergency department with septic shock to receive either fluids at restricted volumes and early vasopressor therapy (vasopressor group) or higher volumes of fluids and later vasopressor therapy (fluids group) for at least 6 hours and up to 24 hours. The primary outcome was days alive and out of the hospital from randomization to day 90.
Results: A total of 1000 patients underwent randomization, with 499 assigned to the vasopressor group and 501 to the fluids group. Informed consent was not obtained for 37 patients, which left 963 patients in the intention-to-treat population (481 in the vasopressor group and 482 in the fluids group). Three patients in the fluids group were lost to follow-up for the primary outcome. In the first 24 hours after randomization, patients in the vasopressor group received less intravenous fluid than those in the fluids group (median difference, −1108 ml; 95% confidence interval [CI], −1395 to −850). The percentage of patients who received vasopressors was higher by 18.9 percentage points (95% CI, 13.3 to 24.5) in the vasopressor group. The median number of days alive and out of the hospital at day 90 was 76 (interquartile range, 55 to 83) in the vasopressor group and 76 (interquartile range, 55 to 82) in the fluids group (difference, 0.0 days; 95% CI, −2.7 to 2.7; P=1.00). Adverse events occurred in similar percentages of patients in the two groups, except for pulmonary edema (0.6% in the vasopressor group vs. 5.0% in the fluids group; P<0.001).
Conclusions: Among adult patients who presented to the emergency department with septic shock, an approach that involved restricted fluid volume and early vasopressors did not result in a greater number of days alive and out of the hospital at day 90 than an approach involving greater fluid volume and later administration of vasopressors. (Funded by the Australian National Health and Medical Council Medical Research Future Fund and the New Zealand Health Research Council; ARISE FLUIDS ClinicalTrials.gov number, NCT04569942.)
What is the clinical problem?
Liberal fluids are pretty ubiquitous in the ED for people with suspected sepsis in my experience. Certainly in the UK. Surviving Sepsis guidelines condone this practice, offering a conditional recommendation to support use of 30ml/kg in initial bolus doses as an acute resuscitation strategy, usually equating to at least 2L before anyone engages their brain further. However, the presenters today rightly highlighted that we don’t actually know if this approach is effective, and we don’t really understand the harms. In fact, we are still relatively unclear about the best early approach to sepsis resuscitation in general, with multiple options all seeming reasonable. There was lots of good chat about the recent studies in this area such as CLASSIC, CLOVERS, PROMISE, ProceSS and ARISE. This concluded with final mention of a recently updated SR/MA with TSA. What was the conclusion? Further trials are needed. Of course. And here is one of them.
Give me an overview of this study
In this study, the investigators set out to evaluate whether a strategy of restricted fluid volume and early vasopressors, as compared with a strategy of greater fluid volume and later vasopressors, would increase the number of days alive and out of the hospital at day 90 among adult patients who presented to the emergency department with early septic shock. All sounds good, but let’s dig into the PICO a little.
Participants
The inclusion criteria required all of
– clinically suspected infection
– SBP <90 or MAP <65 following 1L fluid
– lactate >2
– commencement of antimicrobial treatment
The exclusion criteria were fairly standard (pregnancy, about to die etc..) but the exclusion of patients having already received >2L fluid or >6h since ED presentation are noteworthy. This perhaps speaks to some differences between Australasian acute care systems and others – they can clearly get to their patients very early.
Intervention
The intervention was delivered over the first 24h of care and included multiple components – local investigators were asked to cease fluid resuscitation, start vasopressors, then administer judicious 250ml bolus dose fluid only when deemed necessary. Maintenance fluids were discouraged. Importantly, choice (and route of administration) of vasopressor was left entirely to the treating clinician, as were titration strategies and choice of MAP target. I was particularly interested by this, as I would suggest there are vasopressors and there are vasopressors, and it is important how you use them. More on this later.
Comparator
This was not ‘usual care’ as it so often is in resus trials, but actually fairly didactic. Local teams were advised to deliver IV fluids as the initial resuscitation strategy up to 1L, followed by judicious fluid bolus dosing up to 500ml at a time. Vasopressors could then be commenced if appropriate. Maintenance fluids could be given as per usual practice. This is a reasonable comparator, but risks contamination of groups and thus dilution of treatment effect. Again, more on this later.
Outcome
The investigators selected a primary outcome of days alive and out of hospital up to 90 days (DAOH90). Patients dying before 90 days were assigned zero DAOH. A reasonable patient-centred outcome, but it is a little subjective and can also be affected by multiple therapy components. It also then mandates that you power the study based on potential increase in DAOH, but how many days out of hospital are a good thing?
Methods
The study was set up and delivered across 61 sites, mainly Australasian Emergency Departments. I think there were 5 sites in New Zealand and 1 in Ireland. The consent process was clever, as the authors considered both the intervention and control options to be in keeping with usual care options. As such, they did not require written informed participant consent, but allowed a process of opt out/consent to continue. This will increase numbers, enhance the potential for a consecutive sample and improves the internal validity of the study. The authors published their protocol and statistical analysis plan prior to these results, so the methods can be carefully deconstructed at your leisure.
Randomisation was delivered in a 1:1 permuted block strategy, stratified by site, accessed by internet platform. This should adequately protect allocation concealment. The trial was open label, which of course does risk issues with bias arising from differences in care based on conscious or subconscious choice by clinicians. It also creates the issue of potentially more patients in the vasopressor group getting admitted to critical care, which is a fairly powerful confounder – if some patients are getting 1:1 nursing and twice daily CCU consultant reviews, that is quite different to ongoing care on a medical ward I suspect and could easily influence a 90 day primary outcome centred on hospital length of stay.
No extra resources were given to departments to deliver the study (apart from research nursing support). This design aspect increases the potential generalisability of the findings, provided that busy EDs could deliver both interventions during the study.
The team powered the study to detect a 7 day difference in DAOH. This is arguably a little optimistic and is not really explained in the paper. Why not 3 days? Probably because the sample size would have been unpalatable. The authors highlight the median DAOH of 60 days in the prior ARISE trial, so I suppose they are saying we think a roughly 10% increase in absolute effect would be important, but it would have been good to see a bit more explanation around how they arrived at this choice, and how they included patients and their relatives in the decision making process.
Results
1000 participants were recruited over 4 years, around 50% of those eligible. Usual exclusions are clarified in Figure 1 but centred mainly on too much prior fluid, outside the time frame, or the treating clinician determined the individual was unsuitable for recruitment.
There were 963 participants in the intention to treat analysis. Good balance between groups is shown in table 1 and supports the success of the randomisation process. The average patient had an age in the late 60’s, with a median APACHE score of 18, SOFA score of 4, lactate of just over 3 and got roughly 18mls/kg of fluid prior to randomisation. Most participants had either chest or urinary sepsis.
Figure S5 in the appendix is very interesting and helps us to evaluate a specific concern. Regarding disposition, 61.3% and 72.5% of intervention patients were being managed in an Intensive Care Unit at 6 and 24h respectively, compared to 48.8% and 65.5% in the comparator group. This suggests that people allocated to the vasopressor group were more rapidly referred to critical care, and more commonly managed on critical care, during the first 24h.
What about fidelity to the intervention
I think we need to talk about this separately, as it is absolutely critical to a trial like this. If you allocate heterogenic sepsis patients to either fluids or vasopressors, but then allow delivery of both with a degree of clinical discretion, what do the clinicians actually do? This is essential to consider in detail as it defines the fundamental scientific experiment. If they end up doing roughly the same thing in both groups, then no experiment has taken place – so it would be no surprise if there was no difference between groups.
The authors provide good transparent data on this and a reasonable level of assurance. In terms of vasopressor commencement, 86% of the intervention group received at least 30 minutes of therapy, compared to 67.6% of those in the comparator arm. Vasopressors were started at a median of 21 mins from randomisation in the intervention group, compared to 81mins. Although these are reasonable levels of separation, the use of vasopressors in more than two thirds of the comparator arm, commenced within a median of 90 mins from randomisation, suggests that this was more a trial of restricting fluids, rather than restricting vasopressor use.
Speaking of which, over the 24h intervention period, the intervention group received approximately 1108ml less intravenous fluid, which is a difference of roughly 14ml/kg vs 29ml/kg by the end of the day. Absolute protocol deviation rates were in keeping with other pragmatic drug trials – of around 5% – usually a fluid bolus >250ml in the VP group.
Of interest, there were high rates of central line insertion in both groups (39.1 and 37.3%). Vasopressors were initially delivered via a peripheral route in 71.9% cases in the intervention group and 55.4% in the comparator group. One third of patients got steroids, commenced at a median of 3h. You may say you don’t care about any of this detail but I think you should – if you are looking to apply these results to your population by assuming they are generalisable, then we need to look at exactly what happened to them in general, and whether these things happen in your population.
Ok sure, but what about the key results
The primary outcome was outrageously similar in both groups. DAOH90 for both groups was exactly 76, albeit with different confidence intervals, but with an absolute difference of 0 (95% CI -2.7 to 2.7) days. There were more deaths by both 28 and 90 days in the early vasopressor group.
There were no statistically significant signals in prespecified subgroups, although if you want to talk about ‘trends’ then we could highlight the direction of effect towards favouring early vasopressor use in younger patients, female participants, those with higher APACHE scores, a lactate >3 and those with respiratory infection. Confidence intervals for effect size are very broad for most of these groups
What about signals of harm?
This is also worth digging into, as a lot of the anxiety around delivery of peripheral vasopressors in the UK remains set in the perception of risk from extravasation or mismanagement. Accepting this study reflects delivery in an Australasian system, adverse event rates from vasopressor use were remarkably low. Table S11 highlights a 0.4% risk of extravasation in the early vasopressor group (with 0.0% tissue necrosis). This fits with previous safety data reported in CLOVERS, showing that peripheral pressors were safe with a complication rate of 3%, and all the complications were minor and reversible.
There were no SAEs in either group. The only real safety issue of note was a higher rate of predefined acute cardiogenic pulmonary oedema in the fluid arm, seen at a rate of 3 (0.6%) and 24 participants (5.0) in the intervention and comparator groups respectively (p=0.12 (0.03 to 0.39). Even with this higher incidence, there was no evidence of any increase in organ support requirements such as NIV or effect on the primary outcome.
What about generalisability?
This was a well conducted study delivered in an Emergency Department setting on patients with septic shock. We continue to see these patients, and there is still significant uncertainty about the initial resuscitation treatment option.
However, it is worth noting that almost all patients in this study were recruited in the Southern hemisphere. You might say this is a broadly similar health system in another high income country, so the evidence is directly applicable to your UK/USA patients. But we should remember the recent EGDT trials highlighting an intriguing anomaly in outcomes for septic shock – ProMISe in the UK had a 90 day mortality of 29% for example, whereas ARISE in Australasia had a 90 day mortality of 18%. Why such a big difference in patients with essentially the same disease? Presumably a melting pot of known and unknown confounders, such as differences in case mix, illness severity, deprivation, departmental flow and access to acute resuscitation measures. This area is still a little grey, but does suggest there should be some caution in assuming there is direct readability from ARISE-FLUIDS across to other areas of international practice.
What about strengths and limitations
We should highlight here that CCR always delivers an editorial on each paper presented, so the below is really a distillation of a great presentation by Tine Sylvest Meyhoff from Copenhagen, with some additional thoughts of our own.
First, this was an impressive demonstration of a randomised trial delivered in an acute setting with a clever approach to ensure early approach and randomisation. The investigators also accommodated movement within the hospital (protocol delivered on CCU as well as ED) which is not easy and requires engagement of multiple stakeholders. The study was pragmatic and delivered with a valid outcome important to both patients and clinicians. However, in spite of these strengths, this study did not deliver on a plausible biological hypothesis supported by prior observational work. It is therefore important to think about why that might have occurred. Tine highlighted this excellent NEJM article which makes us perhaps step back to address this issue and ask the following questions:
Was it an appropriate intervention?
Protocolisation of the intervention is a continuing Achilles heel of sepsis trials. Sepsis is a heterogenous condition which can present with hypovolaemic shock, distributive shock, cardiogenic shock or more often a variable combination of all 3. We don’t have a simple way of objectively and rapidly confirming the relative contribution, unfortunately. The idea that in the absence of certainty, you would restrict access to any specific therapy in the context of a clinical trial, is clearly unpalatable to clinicians, and as such to patients.
In this study, the authors went for early inclusion and pragmatism, rather than didactic protocolisation. They allowed autonomy on vasopressor delivery and provided ‘guidance’ on restricting fluids, but mandated little. There did appear to be some separation between the intervention and comparator group by the end of the study, with the VP group receiving more VP, earlier commencement of VP and less fluids by 24h. We need to be absolutely transparent here though – more vasopressors actually meant 19% more participants received vasopressors (86% vs 57%, not 100% vs 0%), earlier commencement meant 60 minutes earlier and less fluids meant 1L less by 24h from randomisation. If the investigators had set these differences out as their proposed intervention at baseline, I think we would probably have been sceptical about the chances of finding a significant effect.
There is also a point to be made about the specifics of vasopressor use. The data on this is slightly buried in the supplementary appendix, but table S6 will tell you that the vasopressors used in the intervention arm included Noradrenaline (47.3%), Metaraminol (25.2%), Vasopressin (5.2%) and Adrenaline (2.5%). I think we would argue that there are vasopressors and there are vasopressors. For example, noradrenaline will have multiple effects on the body, including increasing vasomotor tone through alpha adrenergic stimulation, mobilising unstressed volume through venoconstriction and potential positive inotropy through beta agonism. Metaraminol will likely only provide a degree of alpha agonism and arterial vasoconstriction. This is why it works well in post induction vasoplegia in well patients undergoing anaesthesia, but works less well in the critically ill. By lumping these agents together, there is a very real risk of diluting any treatment effect in the intervention group. Perhaps this contributed to the neutral result.
Was the targeted effect size appropriate?
This is a proxy question really, asking whether it was big enough? The investigators did really well with this study, and it is amongst the largest fluid RCTs in sepsis. They also recruited early, followed up well and reported a tremendous amount of data. They will likely have had sufficient power to reject their original hypothesis, if there was a true difference of 7 DAOH90. But was a 7 day DAOH difference too optimistic, when the actual length of stay in this cohort appeared to be a median of 14 days? Is it reasonable to suggest that early VP for the first 24h will halve the duration of admission, when all care after 24h is likely to be relatively similar between groups? Could they have selected 5 days, or 3 days? Of course they could, but this would likely have required a much larger sample size.
Was the primary outcome an appropriate choice?
This is a good question. You could go for mortality here, or surrogate markers of improvement such as trending SOFA scores or organ support requirements. You could also go for rates of CCU admission, days in CCU, organ support free days or general hospital LOS. However, DAOH90 is a good barometer for all of these and perhaps preferred by patients. 90 days seems a little arbitrary given the intervention was mainly delivered during the first 6-24 of hospital stay. This risks further dilution of treatment effect, given all of the things that can happen subsequent to admission and directly impact discharge decisions. However, this time period for follow up could of course be debated ad infinitum and there is unlikely to be a Goldilocks number of days that everyone will agree is the right number.
The early bird catches the worm
We can’t finish this section without also talking about pre-randomisation fluid. If we are testing the hypothesis that early vasopressors and not much fluid is a better strategy for sepsis, then we need to be sure that people don’t get too much fluid before enrolment. But how much is too much? And can we realistically ban fluid in septic patients while waiting for a research team to attend?
Median IV fluid given prior to randomisation was 1500ml (18ml/kg) in this study, which is a decent amount. Many of you will remember the FEAST study in septic shock showing a mortality benefit with iv fluid restriction, at a level of 10ml/kg. Observational data also shows a dose response curve, suggesting that iv fluid restriction needs to be at a level that would make most clinicians feel really uncomfortable, if it is going to be an effective strategy. This trial mandated a 1000ml iv fluid bolus prior to eligibility check to ensure patients didn’t have pure hypovolaemia.
All very interesting. Any final thoughts?
Despite gestalt, POCUS and various other tools we don’t currently have a way in the ED to separate out the hypovolaemia from the distributive vasoplegia from the cardiac suppression (or the combination) at the front door. These patients are sick and need time critical treatment so everyone is very uncomfortable doing nothing. We need to recruit at the earliest point before any treatment has been given. Other studies on early vasoactive medication in sepsis (such as the NIHR funded EVIS trial) are actively recruiting with this in mind, and it will be interesting to see whether the pre-randomisation fluid delivery is any different.
Well, the authors rightfully conclude that in patients presenting to the emergency department with septic shock, administration of restricted volumes of fluids and early vasopressors, as compared with greater volumes of fluids and later vasopressors, did not increase the days alive and out of the hospital from randomization to day 90.
The editorialist had a slightly different spin, suggesting that you could also draw conclusions along the lines that early vasopressors appear safe, and so do early fluids. Ergo, it doesn’t seem to matter too much what you do. All in all, this is a helpful study and result, as it frees us up to do what we think is best after application of our clinical gestalt, supplemented by point of care assessment and initial investigations. If you think your patient is behind on fluids, you probably won’t do much harm by pushing a fluid strategy. If you think your patient is well filled but vasoplegic, it would appear relatively safe to commence early peripheral vasopressors. You could do both at once.
We think some of the challenges with dilution of treatment effect, optimism with the initial effect size and mixed approaches across the two groups in this study mean that further research is still likely to be helpful and informative.
For now, we should probably reflect on these findings and temper our enthusiasm for peripheral vasopressors. In Virchester, their use mandates admission to a critical care environment. This is a costly undertaking and there is no country in the world that has unlimited critical care capacity. Given the lack of demonstrable benefit in this study and the wider implications for resource use, perhaps we need to be more judicious with their use, while recognising that these data support a low level of adverse events and/or complications.
Thanks all.
Dan and Alasdair (and again, thanks to Tine for an excellent live editorial following the presentation).
DOI – We should declare active research roles in this area – Alasdair is the chief investigator for the NIHR funded Early Vasopressors in Sepsis, and I am co-investigator. The study is currently recruiting.
Key references:
- The ARISE-FLUIDS Investigators. NEJM 2026; Online first DOI: 10.1056/NEJMoa2516225
- Shi et al. Crit Care 2025; May 6;29(1): 182.
- NHLBT ACLI group. N Engl J Med 2023; 388: 499-510
- Mouncey et al. N Engl J Med 2015;372:1301-1311
- ARISE investigators and ANZICS. N Engl J Med 2014;371:1496-1506
- PROCESS investigators. N Engl J Med 2014;370:1683-1693
