(Last Updated On: December 12, 2018)

Balanced Crystalloids versus Saline in Noncritically Ill Adults

WH Self et al. New England Journal of Medicine, March 2018. [paper]

Why I chose this article

This article is another flashpoint in the long debate – typically between internists and surgeons – regarding the use of normal saline versus balanced crystalloids for resuscitation. I remember being exposed to this issue when I was in my general surgery rotation in medical school and overhearing the surgeons chastising non-surgeons about their choice of normal saline (NS) instead of Lactated Ringer’s (LR) for resuscitation. Since then, I’ve always had this nagging feeling that I ought to spend more time looking into this debate and at the pros/cons of both NS and LR.


NS is the default resuscitative fluid but it is not a “normal” physiologic fluid akin to human plasma. It’s a hypertonic solution that causes a hyperchloremic acidosis, which can worsen an acidemia in an ill patient. NS could also cause hyperkalemia secondary to an acidosis. Kidney function can be impaired by NS infusion by precipitating a hyperchloremic state and thereby reducing renal perfusion. We use NS all the time in the emergency department and upstairs on the floor but is this appropriate and safe for our patients? Am I torching kidneys with every clicked order of 0.9% NS 1,000 cc? Am I worsening a tenuous acid-base imbalance with my default order of NS?

There are no large trials examining how resuscitation of NS affects our patients who are admitted to the floors in the long-term. This article seeks to rectify that. (Nota bene: there is an accompanying article in the NEJM that asks this question for critical patients admitted to the ICU from the ED.)

Research Question

Does the composition of crystalloid fluid used in fluid resuscitation – namely normal saline or balanced crystalloids – have an effect on patient outcomes, specifically kidney function and “hospital-free days,” in non-critically ill patients?

For the purpose of this journal club, I’ll be using NS for normal saline and BC for balanced crystalloids, which includes both Lactated Ringer’s and Plasma-Lyte A in this particular trial.

Study Design

The aptly named SALT-ED trial – Saline Against Lactated Ringer’s or Plasma-Lyte in the Emergency Department – was conducted at Vanderbilt. It is a single-center, multiple-crossover trial comparing BC with NS among adults who were treated with intravenous crystalloids in the ED and were subsequently hospitalized outside of an ICU. The trial lasted 16 months and the choice of crystalloid would alternate every month, with the entire department using NS for one month and then switching to BC the following. As such, this trial was unblinded: staff knew which type of solution they would be hanging. Importantly, there were two options for the BC month: LR or Plasma-Lyte A.

Inclusion criteria

  • Adults over the age of 18
  • Received a minimum of 500cc of IVF in the ED
  • Admitted to the hospital but not to the ICU

Exclusion criteria

  • Patients that received less than 500cc of IVF in the ED
  • Patients who were admitted to the ICU

Strangely, people with ESRD on HD were also included though they could not obviously contribute to many of the measured secondary outcomes such as a new need for dialysis, persistent kidney dysfunction, and AKI. They would however be included in the MAKE30 composite depending on whether they died during the trial period.


Primary outcome: “hospital-free days,” or the number of days a patient was alive after discharge up to day 28.
Secondary outcomes: (1) death from any cause, (2) a new need for dialysis, or (3) persistent kidney dysfunction, which was defined as an elevation of SCr ≥200%. These were measured and then compiled into an unusual composite called Major Adverse Kidney Events at 30d (MAKE30). These outcomes also consisted of any new diagnosis of AKI of stage 2 or higher and any in-hospital mortality. These outcomes were assessed at hospital discharge or at 30 days, whichever occurred first.


Intention-to-treat analysis was used for all eligible patients, so every patient that met inclusion criteria and received either NS or BC was included in the final analysis. The primary variable, hospital-free days, was analyzed with a multivariable proportional-odds model; this is a variant of a logistic regression model but uses dichotomous dependent variables. Secondary variables were analyzed using multivariable logistic-regression models and adjusted for age, sex, race, admitting inpatient service, and days that had elapsed since the beginning of the trial.


  • 13,347 patients were enrolled: 6708 (50.3%) received BC and 6639 (49.7%) received NS.
  • The majority of these patients were white, with 5159 (76.9%) assigned to BC and 5189 (78.2%) assigned to NS.
  • Most of these patients were hospitalized under a medicine service – around 70% for both groups; the remaining 20% or so were hospitalized under a surgery service.
  • Around 32% of patients in each group received more than 2L of crystalloid.

One of the biggest takeaways from the study is that there was no difference in the number of hospital-free days between the BC and NS groups: each group had a median of 25 days (adjusted OR with BC, 0.98; 95% CI, 0.92 to 1.04; P=0.41). The second biggest take-away is that patients who arrived at the ED with underlying renal dysfunction – defined as sCr >1.5 mg/dL – or hyperchloremia (110>mmol/L) were least likely to suffer from the composite outcome MAKE30 and AKI when receiving BC. Compared to the NS group, patients receiving BC had a lower incidence of MAKE30 within 30 days (4.7% vs. 5.6%; adjusted OR, 0.82; 95% CI, 0.70 to 0.95; P=0.01). However, when looking at each of the secondary variables on their own merit, no statistical significance was reached when comparing BC with NS. This difference in MAKE30 between the two groups is driven by the increase in sCr in the NS group.

Great wave off the coast of Kanagawa by Katsushika Hokusai, c. 1830.


In non-critically ill patients, whether you use BC or NS, it won’t get the patient discharged from the hospital any sooner and it won’t flame out their kidneys. However, for patients coming in with baseline renal dysfunction, using LR may result in fewer instances of AKI and adverse effects. Sounds simple but there is a lot to unpack with this trial.

There are many strengths to this clinical trial. This is the first of its kind that looks at patients who have come through the ED and have been subsequently admitted to the floor, under the care of both medicine and surgery services. It was a huge endeavor with over 13,000 enrolled patients who appear to have been appropriately randomized and well-matched. Of those enrolled, more than 85% received the crystalloid du jour or du mois. It was also conducted at the same time as the SMART trial examining the very same research question but for patients admitted from the ED to five ICUs. But I do have a lot of reservations about this article and its methodology.

The trial was conducted at one medical center and over 76% of patients were white. How generalizable are these results to most inner-city populations and to our patients here in Chicago?

This MAKE30 composite variable is bizarre. Why lump together death, new need for dialysis, and doubling of creatinine? These are not synonymous clinical outcomes in any setting. It seems to be an effort of statistical hijinks by the authors to smush some variables together and come up with at least one statistical difference between NS and BC. I’m not convinced that this is a useful variable for distinguishing these two groups.

As an unblinded study, the potential for selection bias is high. Did providers change their level of care based on what type of crystalloid they could infuse? I feel that this could have somehow been manipulated in the ED to allow for formal blinding of physicians and enhance the integrity of the trial.

I’m not thrilled that both LR and Plasma-Lyte were included in the balanced crystalloids option. Plasma-Lyte has an inherently different composition from LR: it contains magnesium, acetate and gluconate, and has a higher osmolarity that actually approaches NS and human plasma (Plasma-Lyte 295, LR 274, NS 308). Why not just stick solely with LR? Especially as Plasma-Lyte doesn’t even appear to be a go-to option in this hospital (95.3% of the time LR was used). Did the choice of two options for the BC group dilute – pun intended – the results?

I’m also troubled by the fact that the choice of crystalloid administered in the ED was not carried through while on the floor – “fluids administered after hospital admission and those used as medication carriers were not controlled.” So we have no idea what was infused nor how much of it once they were admitted to the floor. Not only that, but I’m sure that for patients who received IVF in an ambulance en route to the ED received was NS, which further confounds their results.

Other lingering questions that I have: Why is the primary outcome measured up to day 28, with secondary outcomes at day 30? Does it matter what these patients were admitted for? For 35% of these patients, a baseline Cr is unknown and was calculated. How does this affect the results and does it matter?

Will this change my practice? I was hoping it would when I saw the cover of the NEJM on my doorstep, but for now I’m going to continue to use NS as my go-to resuscitative solution. I’ll happily consider using LR for patients who come into the ED with a sCr >1.5. I know NS is not “normal” and the arguments for avoiding it make good physiological sense but I don’t think this SALT-ED trial presents a vigorous enough argument for dropping it. I need to see this trial replicated and preferably blinded in a multi-center setting. What do my colleagues think?


Rebecca Kreston is an EM/IM Resident, class of 2022



EM/IM Sessions are reviewed in journal club style by the current attendings and residents, as well as alumni of the UIC IM/EM program prior to publication. This post was specifically reviewed by Wes Eilbert, MD, Associate Professor of EM. Elspeth Pearce, MD Editor.