Liberal or restrictive fluids? This RCT found no benefit or harm in sepsis-induced hypotension
Most septic patients don't have a salt water deficit, and some people are (incorrectly) convinced that normal saline is a murder weapon. (BaSICS Trial. JAMA. 2021;326:818; https://bit.ly/3Y06Yqj.)
Should salt water remain the first-line therapy for sepsis-induced hypotension? The CLOVERS trial, just published, suggests that perhaps nothing we do matters. (N Engl J Med. 2023 Jan 21. doi: 10.1056/NEJMoa2212663. https://bit.ly/3WH6KmG.)
The CLOVERS trial was a multicenter, open-label, randomized, superiority trial using a convenience sample of adults with suspected or confirmed infection (defined as having received antibiotics) and sepsis-induced hypotension (systolic blood pressure less than 100 mm Hg after a 1 L bolus).
Patients were excluded if it had been more than four hours since hypotension began, more than 24 hours since hospital admission, more than 3 L of IV fluids had been given before enrollment, or they had fluid overload or volume depletion from a nonsepsis source.
The trial used a restrictive fluid strategy that prioritized vasopressors as the primary treatment for sepsis-induced hypotension, with “rescue fluids” being permitted for prespecified indications that suggested severe intravascular volume depletion, and a liberal fluid strategy consisting of a recommended initial 2000 mL intravenous infusion of isotonic crystalloid, followed by fluid boluses administered on the basis of clinical triggers (e.g., tachycardia) with “rescue vasopressors” permitted for prespecified indication.
No Statistical Differences
The primary outcome was all-cause mortality by discharge or 90 days. The researchers enrolled 1563 patients from 60 hospitals in the United States. Patients appeared evenly matched at baseline, with a mean age of about 60, a SOFA score of 3.4, and a mean systolic blood pressure of 93 mm Hg. Patients had received a median of 2 liters of IV fluids prior to randomization in both groups, and about 20 percent of each group was on vasopressors at the time they were enrolled.
Protocol adherence was 97 percent, and the treatment difference between the groups appeared clinically significant. Median fluids in the first six hours was 500 mL in the restrictive group compared with 2300 mL in the liberal group (-1800 difference, 95% CI -1889 to -1711). The median amount of IV fluid was 1267 mL in the restrictive group compared with 3400 mL in the liberal group over 24 hours. Vasopressor use was 59 percent in the restrictive group compared with 37 percent in the liberal group.
No difference was seen in the primary outcome of all-cause mortality (14.0% vs. 14.9%, ARR 0.9%, 95% CI -4.4 to 2.6%, p=0.61).
There were no statistical differences in any of the secondary outcomes (nor do there appear to be any underpowered clinically important differences), including ventilator use, renal replacement therapy, ICU length of stay, hospital length of stay, or serious adverse events. The subgroups also all look identical (although perhaps there is a hypothesis generating point estimate in the group of patients with end-stage renal disease).
These results could make you nihilistic, but I don't think there is any doubt that fluid and vasopressor support really matter for some septic patients. What I think this trial really emphasizes are the limitations of our definitions in sepsis (we group together a lot of disparate conditions under a single umbrella), the limitations of our current monitoring (we don't know if these patients had sepsis-induced cardiac dysfunction), and the limitations of using blood pressure as a surrogate (when we really care about tissue perfusion).
There is almost certainly a subset of patients where hemodynamic support really matters, but this trial (among many others) makes it clear that the decisions aren't critical for the average patient. Future research really needs to delve into the heterogeneity of sepsis, rather than treating it as a homogenous state of pathology.
This is a large trial, but it is still not large enough. The 95% confidence intervals include numbers that we would care about on both sides of the coin (from restrictive fluids decreasing mortality by 4.4% to restrictive fluids increasing mortality by 2.6%). There is no reason to think that either of those outcomes is real, but they are within the realm of statistical possibilities after this trial.
Crossover is always a problem in these kinds of trials, and it can bias the results toward no effect. The restrictive group still received a fair amount of fluid, including everyone getting about 2 L IV before being enrolled. This is probably not an ideal study from the standpoint of some people in the restrictive camp, but I think it is pragmatic because most patients will get a few boluses as we sort out what is going on in the emergency department. Compared with other trials we have seen, there is actually a pretty big gap between the groups, and I think the difference of 2 L over a 24-hour period is clinically significant and large enough that you might have expected a clinical difference in some of the outcomes (even if you don't think 2 L of salt water is going to change mortality).
One could imagine that the inclusion criteria for this trial were imperfect. Perhaps patients should have been selected based on some kind of volume status assessment. That being said, I don't think that anyone has proven the clinical utility of volume assessments, so who knows if that would have made a difference?
It is also possible that the inclusion criteria encompassed too many “healthy” sepsis patients. Only 20 percent were on vasopressors at the beginning of the trial, and only 59 percent of the restrictive group ended up on vasopressor at any point. The mortality was about half that seen in the early goal-directed therapy trials (although mortality rates for sepsis are lower across the board compared with 20 years ago). The inclusion of too many “healthy” sepsis patients could dilute an effect (in either direction) in the sicker subset.
I think there are some significant practical advantages of using the restrictive algorithm that wouldn't necessarily be captured in a study. Despite decades of attention, septic patients with borderline blood pressures are still routinely admitted to ward beds with limited monitoring, and they deteriorate without anyone being aware. The brief blood pressure response to a fluid bolus can also be misleading and frequently lead to patients being admitted to a floor with a lower level of care than they really require.
Earlier use of vasopressors in sick patients prevents that from happening because almost all hospitals require patients on vasopressors to be admitted to the ICU (which is where almost all of these patients deserve to be). That difference wouldn't necessarily be identified in a trial like this because patients often receive above-average monitoring in a research context. Therefore, I still lean toward early vasopressors in patients about whom I am concerned.
As a side note, only 27 percent of the restrictive group had a central line placed within the first 72 hours, despite 59 percent receiving vasopressors. Forty percent received peripheral vasopressors at some point, but it seems that almost 30 percent received their vasopressors exclusively through a peripheral IV for their entire ICU stay.
No infusion site extravasations were reported in this study, and the only skin complications occurred in the liberal fluid group. In other words, as we already knew, peripheral vasopressors are safe and clearly part of the standard of care. (First10EM. Dec. 11, 2019; http://bit.ly/3JcncZd.)
The bottom line: The CLOVERS trials is an open label RCT that did not demonstrate any difference (beneficial or harmful) from using a restrictive fluid strategy over a liberal fluid strategy in sepsis-induced hypotension. Do whatever you please for now, although for practical reasons I am still leaning toward earlier vasopressors in sepsis.
This article first appeared in Dr. Morgenstern's blog, First10EM (Jan. 24, 2023;http://bit.ly/3kD3snf).
Dr. Morgensternis a community emergency physician practicing in Toronto, Canada. His blog,https://First10EM.com, is dedicated to resuscitation and evidence-based medicine and encourages skepticism and a mindset of constant questioning, humility, and scientific reasoning in medicine. Follow him on Twitter@First10EM.