Authors: Rob Hirst, Chris Connolly, Becky Maxwell / Codes: IP3, RP6, SLO3 / Published: 06/03/2023

Today we are going to be talking about the CLOVERS Trial (cue Crimson & Clover).

The CLOVERS trial stands for: Crystalloid Liberal Or Vasopressors Early Resuscitation in Sepsis from the PETAL network – PETAL stands for Prevention and Early Treatment of Acute Lung Injury.

This study is looking to examine whether a restrictive or liberal approach to fluid resuscitation before vasopressors improves outcomes. There have been a number of studies looking to answer this question, including the CLASSIC study, and the EVIS study which is a UK study that is currently ongoing. Many previous studies have evaluated standard versus restrictive fluid protocols in ICU patients, but this trial is the first to look at mortality difference in patients who almost exclusively presented to the emergency department with sepsis.

So, let’s look at this study.


This study was a multi-centre, unblinded randomised controlled trial at 60 US EDs. Patients were randomised 1:1 to a restrictive or liberal approach to intravenous fluid therapy using a centralised system and stratified according to trial site.

Patients were analysed using an intention to treat analysis and the primary outcome was death from any cause before discharge home or at 90 days. The trial had pre-specified stopping boundaries at 1/3 or 2/3 recruitment and there was a monitoring board to ensure this was appropriate.


Included patients

Adult patients who presented with suspected sepsis, which was defined as the administration or planned administration of antibiotic agents with sepsis induced hypotension – this was defined as a systolic blood pressure below 100 after receiving 1L of intravenous fluids. Patients were recruited on a convenience basis, with less coverage at weekends and overnight.

Exclusion criteria

Patients who experienced a delay of >4 hours since meeting the criteria before they were randomized, >24 hours since presentation to hospital, having received 3L of IV fluids prior to randomisation (including in the pre-hospital stage), presence of fluid overload or severe volume depletion from non-sepsis causes.


There were 12276 patients screened for eligibility, around 4868 met eligibility criteria and 1563 were randomised 1:1 to each of the two groups. Their baseline characteristics were well matched in terms of age, gender, co-morbidities, SOFA score (an average of 3.4 in both), blood pressure, lactate and mean volume of fluid administered before randomisation – exactly 2050mls in both.


The trial followed all patients for 24 hours post-randomisation, with hourly reassessments or after an intervention. This next bit is a bit wordy so I apologise for that!

Restrictive group

The restrictive group prioritised vasopressor administration. If the patients’ MAP or systolic blood pressure dropped below 65 or 90 respectively patients would receive norepinephrine to achieve these targets. A second vasopressor could be added if necessary.

Whilst all maintenance and bolus fluids were halted, patients could receive rescue boluses of up to 2L if they met criteria such as severe or refractory hypotension, a lactate >4 or a rising lactate despite therapy, a HR or >130 for 15 minutes or echo or haemodynamic evidence of extreme hypovolemia. Administration of rescue fluids could be given at any time by the clinical team at their discretion.

Liberal group

All maintenance fluids were stopped. Patients were given a 2L bolus at randomization, although a protocol amendment added a year into the study also allowed patients to stop after 1L if clinician determined the patient was volume replete. 500ml boluses of fluids were administered if they met criteria such as MAP or SBP below 65 or 90 respectively, a lactate >4, decreased urine output (<30ml/hr), a HR > 110bpm or receipt of vasopressors to maintain BP.  ‘Rescue vasopressors’ could be administered if they met criteria of severe or refractory hypotension, a lactate >4 and increasing after 2 hours of therapy, clinical manifestations of fluid overload or a total of >5L administered. Administration could be given at any time by the clinical team at their discretion.


There was no difference in the primary outcome of death before discharge home by day 90 – around 14% in both groups, and there were also no differences in any of the secondary outcomes, which included organ support, ventilator use, renal replacement therapy, ICU length of stay, hospital length of stay or serious adverse events. Of note, looking at the subgroup analyses patients with end-stage replacement therapy may benefit from the more liberal approach to fluid resuscitation, but this is purely hypothesis generating at this stage.


So, what’s good about this study?

A lot! This study reflects the nature of our work and the broad inclusion criteria in the ED mean results are well-suited to answer this question we’ll encounter in our regular practice.

It had an excellent design with a high degree of internal validity

There was a very high level of adherence to the protocol – the first 300 patients and 10% of the patients had were audited and found to adhere ~97% of the time in both groups.

The patients achieved a good degree of separation in terms of fluids administered, which is what you’d want in this sort of trial. The restrictive group received a median of 1267ml in the first 24 hours and the liberal group a median of 3400ml, even allowing for the fact that both sets of patients received 2 litres in the first instance before randomisation.

Great safety outcome reporting, including on peripheral vasopressor administration. In order to ensure patients could access vasopressors early they focused on peripheral vasopressor administration, and from 500 patients treated using peripheral vasopressors there were only 3 cases of extravasation, none of which required intervention or resulted in clinical consequence, providing data supportive of this practice.


Now, primarily this may demonstrate some of our shortcomings with the definition of sepsis – we group together a lot of conditions under this same umbrella, which may have different responses to fluids. The use of blood pressure as a surrogate for tissue perfusion may also be inexact. These patients were also probably more well than in other sepsis trials – only 20% were on vasopressors at the beginning of the trial, and just under two thirds of the restrictive group ended up on a vasopressor at any point. The inclusion of these ‘healthier’ patients may dilute the effects in a sicker subset.

The study was obviously unblinded, and there were a few areas where bias could have crept in. Most of the rescue interventions, whether that be fluids or vasopressors were up to the treating clinician, which could introduced bias. There were large numbers of patients eligible but not included – only about ~25% of those screened were recruited, which may introduced a degree of selection bias, including 873 patients excluded due to practitioner refusal.

The study was stopped short due to futility, that is, there was no differences between the groups, but it did mean that some groups, such as the end-stage renal failure group could have reached significance with more patients.

Clinicians did not use dynamic measures of fluid responsiveness to guide fluid administration, and the fluid regimens did not account for weight / size of the patient – 2L in a 50kilo patient is quite different to a 120kg patient.

Trial protocol only went for first 24 hours.

And whilst this was a very large trial, there were still numbers we would want to know about on either side of the 95% confidence interval, from restrictive fluids reducing mortality by 4.4% to restrictive fluids increasing mortality by 2.6%.

However, that said, this study does provide evidence that either approach is probably reasonable in the first instance for the average patient presenting with sepsis to the ED, which is probably what we’re most interested in. So, this won’t be changing my practice, but it will probably allow me to breathe a bit easier and continue managing patients as you would do. It also may provide another bit of information suggesting that peripheral vasopressors is probably a safe and effective strategy providing you’re using a big cannula not in someone’s ACF and you monitor the cannula site.