Saline vs. balanced solutions has been a topic of ongoing debate. Two fresh studies will illuminate this: the SMART and SALT-ED trials. This post summarizes current knowledge, beginning with physiology and working our way to fresh trials.
Reason #1. There is no physiologic rationale for using “normal” saline (NS).
Saline is a hypertonic, acidotic fluid (it's not “normal”)(1). There is no physiologic reason that infusing it into people is a smart idea. Rather than any scientific rationale, the reasons for using saline boil down to the following:
- It’s cheap.
- It’s what we have always done.
- Everyone in the department uses it.
- It's how I was trained.
- It seems to work.
- The nurses start it automatically.
- The cost difference between a liter of saline and liter of lactated ringers (LR) is about 25 cents.
- Saline is historically the most widely used fluid, but that doesn’t prove it’s the best. This belief reflects status quo bias – assuming that conventional therapy is superior.
Reason #2. Saline exacerbates acidosis
Saline causes hyperchloremic acidosis. Some patients we resuscitate are severely acidemic to begin with. It is irresponsible to give them a fluid which will exacerbate their acidemia. If saline is used as the default resuscitative fluid, then it will be given to acidemic patients before their labs return.
Healthcare systems should be designed to avoid causing harm. Using saline as a default fluid is a poor choice, because it will inevitably lead to a situation where a severely acidemic patient is given an acidic fluid, causing deterioration. This is a failing at the system level. The fix is to use a balanced crystalloid as the default resuscitative fluid. Balanced fluids will always pull the patient's pH towards normalcy, so they are safe regardless of the patient's electrolytes (2).
Reason #3. Saline may be dangerous in hyperkalemia
Evidence suggests the exact opposite of traditional dogma: saline may be dangerous in hyperkalemia. The reason for this is that saline causes acidosis, causing potassium to shift into the blood. Four RCTs performed during renal transplant surgery all showed that saline causes more frequent episodes of hyperkalemia (compared to either LR or Plasmalyte) (O’Malley 2005, Khajavi 2008, Modi 2012, Weinberg 2017).
It’s unclear how generalizable this is. From a physiologic standpoint, it ought to be universal. For example, Martini 2013 found similar results when comparing LR vs. NR in a pig model of hemorrhagic shock. However, patients with adequate renal function may be able to compensate for potassium entering the blood.
Reason #4. Hyperchloremic acidosis causes harm in animal models
It's indisputable that saline causes hyperchloremic acidosis. However, the effects of mild hyperchloremic acidosis are debated. Is this clinically significant, or just an abnormal number?
Kellum 2004 evaluated the effect of hyperchloremic acidosis in septic rats. Acidosis was produced by infusing hydrochloric acid at two different concentrations, compared to an equal volume of LR as the control. Acid infusion decreased blood pressure, which isn't too surprising (figure below). What is remarkable is that hemodynamic dysfunction appeared when the severity of acidosis was mild (base deficit ~5 mM). This suggests that in the context of sepsis, even mild hyperchloremic acidosis is harmful.
Orbegozo 2016 performed a study comparing NS vs. LR vs. Plasmalyte in a sheep model of septic shock. Sheep treated with NS fared poorly:
- Greater acidosis
- Reduced cardiac index
- Reduced muscle tissue oxygen levels
- Impaired microcirculatory perfusion
- More coagulopathy
- Earlier development of oliguria
- Earlier death
This study proves that hyperchloremic acidosis due to saline can cause harm. Differences emerged when electrolytes were only slightly altered, again suggesting that mild hyperchloremic acidosis may be dangerous.
Reason #5. Saline may cause hemodynamic instability
Potura 2015 performed a RCT on patients undergoing renal transplantation randomized to receive saline vs. Elomel Isoton solution (an acetate-buffered balanced crystalloid)(3). Patients receiving saline required vasopressors more frequently (30% vs. 15%, p=0.03). However, this was a secondary endpoint.
Pfortmueller 2018 investigated this finding further with a double-blind RCT comparing normal saline vs. Elomel Isoton solution in patients undergoing major abdominal surgery. Patients receiving saline required vasopressors more frequently (the primary endpoint, see below). Both groups of patients received about three liters of fluid, producing moderate differences in electrolyte levels:
This study is limited because it was halted early due to harm, after enrolling only sixty patients (4). Nonetheless, it is quite consistent with Potura 2015 and animal data shown above.
Reason #6. Saline may lead to increased levels of inflammation
Kellum 2006 showed that in septic rats, using hydrochloric acid to induce a hyperchloremic acidosis increased levels of pro-inflammatory cytokines. Similarly, Zhou 2014 found that in a rat sepsis model, resuscitation with saline produced higher levels of interleukin-6, compared to resuscitation with Plasmalyte.
Wu 2011 performed a RCT of patients with pancreatitis that included a comparison between LR and NS. Patients treated with NS had higher levels of C-reactive protein one day after the initiation of resuscitation (5). Normal saline also seemed to delay resolution of clinical features of systemic inflammation (SIRS criteria). Based on this study, LR is generally recommended as the fluid of choice for pancreatitis resuscitation (e.g., American College of Gastroenterology 2013 guidelines).
de-Madaria 2018 replicated the results of Wu 2011. Forty patients with acute pancreatitis were randomized to receive NS vs. LR resuscitation. Patients treated with NS had a nonsignificant trend towards more SIRS criteria (p=0.06). NS significantly increased the level of C-reactive protein (figure above). It remains unclear how generalizable this is to other disease states (e.g. sepsis is similar to pancreatitis).
Reason #7. Saline causes kidney injury
Hyperchloremia causes renal vasoconstriction, so saline is expected to impair renal function. This was indeed found in a rat sepsis model, where saline resuscitation increased the rate of kidney injury compared to Plasmalyte (Zhou 2014). In normal human volunteers, a two-liter saline bolus reduced blood flow to the renal cortex, compared to two liters of Plasmalyte (Chowdhury 2012).
Yunos 2012 performed a landmark before-after study, which involved a structured intervention to avoid chloride-rich resuscitative fluids in an ICU. Following this intervention, the rate of acute kidney injury decreased. Meta-analysis including this trial, small RCTs, and retrospective studies supported the concept that saline causes renal failure:
The SALT-ED trial is a pragmatic cluster-randomized multiple-crossover study performed at the Vanderbilt emergency department. The default resuscitation fluid was alternated monthly between saline and balanced crystalloids (LR or Plasmalyte). This study involved patients being admitted to the hospital ward. A secondary endpoint was adverse kidney events: a composite of death, dialysis, or doubling of the creatinine within one month.
The study included 13,347 patients, who received a median of one liter of fluid. 95% of patients randomized to receive balanced crystalloid were treated with LR, so this was effectively a trial of saline vs. LR. Saline increased the risk of death or renal failure (5.6% vs. 4.7%, p=0.02), with the difference driven mostly by differences in renal failure. As might be expected, the subgroup of patients with renal injury at admission was more susceptible to harm from saline (37.6% vs. 28% adverse kidney events, p<0.001).
The SALT-ED trial confirms that saline increases the risk of renal failure. This trial is remarkable because saline was harmful even at a low dose, in stable patients. Due to the massive size of this study it was able to detect a small difference in renal failure, which would have been missed by most studies.
In most situations, an effect size of 1% wouldn’t be impressive. However, switching from saline to LR is easy to do, at nearly no cost (~25 cents difference). The number needed to harm from saline is 111. This means that the cost of preventing one episode of renal failure is roughly $28. Given that an episode of acute kidney injury is estimated to cost $1700, switching from saline to LR would improve patient outcomes and save money (Silver 2017).
Reason #8. The SMART trial
The research group at Vanderbilt also performed a sister trial involving critically ill patients, the SMART trial. The default fluid used in both the emergency department and the ICU was alternated monthly between saline and balanced crystalloid. Fluid choice was determined immediately in the emergency department and then continued in the ICU (unlike most ICU trials, which ignore initial fluid resuscitation in the emergency department). Adverse kidney events were the primary endpoint (a composite of death, dialysis, or doubling of the creatinine within one month). 15,802 patients were included, with the median fluid intake of 1 liter. Saline was again shown to increase rates of death or renal injury (15.4% vs. 14.3%, with p=0.04).
Perhaps the most interesting aspect of this study is that the difference in the composite outcome was driven largely by a mortality difference (11.1% vs. 10.3%, p=0.06). This suggests that saline isn’t only causing renal failure but may be causing harm via additional mechanisms (e.g. hemodynamic instability and increased inflammation)(6).
Subgroup analysis suggests that saline increases mortality among patients with sepsis (29.4% vs. 25.2%; p=0.02) or chronic dialysis (18.4% vs. 12.2%; p=0.01)(7). This is hypothesis-generating only, but it makes sense:
- Septic patients may be more vulnerable to hypotensive and pro-inflammatory effects of saline.
- Dialysis patients have no endogenous ability to defend themselves from hyperchloremic acidosis.
Reason #9. You have better options
LR is feared for many reasons, most of which are nonsense:
- Hyperkalemia is not a contraindication to LR.
- LR usually has minimal effect on serum lactate levels, so it won’t impair your ability to measure serum lactate levels.
- Cirrhosis or hepatic injury aren’t contraindications to LR (the liver needs to be really dead before it stops metabolizing lactate). Even if the liver were totally unable to metabolize lactate, LR would still be safe. LR can never cause lactic acidosis, because it contains sodium lactate (not lactic acid). Sodium lactate functions as a metabolic fuel for the brain and heart, so if sodium lactate accumulated that would be potentially beneficial.
- Moderate neurologic injury: LR is slightly hypotonic, so theoretically it could increase intracranial pressure. However, rapid administration of large volumes of LR causes only small, transient changes in serum osmolality (table below). Thus, LR is probably fine for patients with mild or moderate neurologic injury (8).
Plasmalyte and LR are both improvements over saline. However, balanced crystalloids aren't all created equal. Currently there isn't any firm evidence regarding the comparison of LR vs. Plasmalyte. My preference is LR for the following reasons:
- LR uses lactate as an anion, compared to Plasmalyte which uses acetate and gluconate. Lactate is a more physiologic choice, which may improve cardiac function.
- Inclusion of calcium in LR might produce a slight hemodynamic advantage.
- The SALT-ED trial provides good evidence for the superiority of LR over saline. Currently, this gives LR the strongest evidentiary support of any balanced crystalloid.
My approach is therefore:
- “Normal” saline is a hypertonic, acidotic fluid. There is no physiologic rationale for its use as a resuscitative fluid.
- There are many potential problems related to saline. These include causing hyperchloremic acidosis, hyperkalemia, hemodynamic instability, renal malperfusion, systemic inflammation, and hypotension.
- Numerous small RCTs have emerged over the past few years which highlight various problems with saline resuscitation.
- The SMART and SALT-ED trials are massive, multiple-crossover trials which compare saline versus balanced crystalloids among critically ill and non-critical patients. Both studies found a 1% increase in death or renal failure with the use of saline.
- The combination of physiology, animal studies, numerous RCTs, SMART, and SALT-ED indicate that it's time to stop resuscitating with saline.
- First principles: pH & balanced crystalloids
- Getting into the weeds: Different flavors of balanced crystalloid
- Physiologically, “normal” is defined as an isotonic fluid, something with the same tonicity as the human body. Since 0.9% saline is hypertonic, it is not actually a “normal” fluid. I was tempted to refer to it as Abnormal Saline Solution (ASS) throughout this post, which would be a more physiologically accurate description.
- For example, if LR is given to a patient with metabolic acidosis, it will tend to increase the patient's pH toward normal. Alternatively, if LR is given to a patient with metabolic alkalosis, it will tend to decrease the patient's pH towards normal. LR isn't always the ideal fluid to fix a patient's acid-base abnormality. However, LR will always be safe, with a tendency to pull the pH towards normal.
- The balanced crystalloid they used was Elomel Isoton, a solution with Na+ 140 mM, Cl- 108 mM, K+ 5 mM, Ca++ 2.5 mM, Mg++ 1.5 mM, and acetate 45 mM. This solution has a higher concentration of biologically active alkali compared to LR or Plasmalyte (which have ~28 mM of either lactate or acetate). Therefore, Elomel Isoton would be expected to be a bit more alkalinizing than LR or Plasmalyte.
- It’s also unclear whether the hemodynamic improvement could relate to some extent to the calcium concentration of Elomel Isoton, rather than solely the pH effects. One way or the other, this study would still seem to suggest that LR is a superior fluid compared to NS (given that LR has a calcium content and pH effects similar to Elomel Isoton).
- C-reactive protein is an accepted prognostic factor in pancreatitis, which predicts outcomes. Levels of C-reactive protein may most closely reflect levels of the cytokine interleukin-6. Therefore, this result in humans is quite consistent with the animal study by Zhou 2014 showing that saline increased interleukin-6 levels.
- Most patients who get acute kidney injury won't be killed by it. Therefore, if saline were harmful solely due to nephrotoxicity, you would expect the number of patients who develop kidney injury to outnumber the increase in mortality. The fact that excess mortality is greater than the frequency of acute kidney injury suggests that saline is causing injury via multiple mechanisms.
- Please note that patients on dialysis could not develop renal failure, so differences in the primary endpoint among dialysis patients are driven purely by a mortality difference.
- Another way to conceptualize this is as follows. LR has an osmolality of 254 mOsm/L and an osmolarity of 274 mOsm. The discrepancy occurs because the ions form complexes that stick together within the LR solution. It’s debatable which figure is the most biologically meaningful, since the ions may dissociate within the blood. To be conservative, let’s use the 254 mOsm/L figure. Using this figure, if we give a patient two liters of LR, that will create a deficiency of 72 mOsm, which is roughly equivalent to giving 250 ml of free water. Now, 250 ml of free water is a quantity of water which we often give without much thought (e.g. it’s in a 250 ml bag of azithromycin formulated in D5W). Therefore, giving a patient two liters of LR should have a similar effect compared to giving two liters of Plasmalyte plus 500 mg IV azithromycin. Nobody would think twice if you gave azithromycin to a patient with elevated ICP.