Considering the importance of crystalloid in critical care, one might expect crystalloid composition to be meticulously engineered and updated. However, our crystalloid choices remain archaic. Normal saline and Lactated Ringers (LR) were developed in the 1800s, whereas Plasmalyte and Normosol emerged in the 1970s.
This post explores how a new crystalloid could be designed, based on modern critical care concepts. The point isn't necessarily the precise formulation of the proposed crystalloid, but rather that it's time to consider something new. Continuing to debate the merits of NS vs. LR or Plasmalyte may be repeating the wrong question.
Problems with existing crystalloids
All available crystalloids have some drawbacks:
Normal saline: This acidotic, hyperchloremic, and hypertonic fluid is the least physiologic. Although believed to be safe in hyperkalemia, NS has actually been shown to increase potassium levels due to intracellular shifts induced by acidosis. The ongoing use of normal saline exemplifies status quo bias. Of course, normal saline remains perfect for a couple of patients a year: patients with hyponatremia, hypokalemia, and metabolic alkalosis due to profound vomiting:
Plasmalyte & Normosol: The drawback of these solutions is their use of gluconate and acetate:
- Gluconate: There is little rationale for including gluconate in intravenous fluid. Gluconate doesn’t appear to be metabolized by the human body, but instead appears to be excreted unchanged in the urine. It might even act as an osmotic diuretic.
- Acetate: Unlike gluconate, acetate is metabolized to yield bicarbonate, so it functions as the alkali in Plasmalyte and Normosol. However, acetate is a poor alkali. Acetate was previously used as an alkali in dialysis fluid, but this was discontinued because it caused hypotension. In an animal model of hemorrhagic shock, infusion of plasmalyte increased mortality compared to LR (p = 0.02; Traverso 1986)(1):
the pro-inflammatory, vasodilatory, myocardial depressant, and hypoxemia promoting properties of acetate led to its removal from contemporary renal replacement fluids. –Davies 2011
Lactated Ringers (LR): This is generally my preferred crystalloid, but it has some drawbacks. It is hypotonic, which could be problematic in patients with elevated intracranial pressure. The inclusion of calcium limits its compatibility with blood transfusion and certain medications.
Foundational principle for designing crystalloid: The target principle
Giving any crystalloid will tend to pull the serum chemistries toward the chemical properties of the crystalloid. For example, consider the effect of LR on potassium. LR has a potassium concentration of 4 mM. Therefore, if LR is given to a patient with a hypokalemia (i.e. 2 mM), it will tend to increase the potassium. Alternatively, if LR is given to a patient with a hyperkalemia (i.e. 6 mM), it will tend to decrease the potassium. Therefore, whatever potassium abnormality may exist (hyperkalemia or hypokalemia), the LR will tend to correct it (2).
In general, when we are constructing a general-purpose crystalloid it makes sense to design the crystalloid to match our target serum concentrations of various ions. Thus, whatever abnormality the patient may have, crystalloid infusion will pull them closer to target serum concentrations. This allows us to safely infuse such fluid into patients whose chemistries we don’t know: regardless of the patient’s derangement, our crystalloid will pull them closer to target. This sort of general-purpose crystalloid might not be the ideal approach to fix the problem, but at least we can be confident that it won’t exacerbate it (3).
Properties of an ideal crystalloid?
An ideal crystalloid should probably be isotonic. This would allow bolusing in patients with increased intracranial pressure without concerns regarding hypotonicity. Additionally, this would tend to pull the tonicity towards normal, based on the target principle.
2. Sodium lactate used as an anion
Historically lactate has been feared, because it is often associated with badness. However, the lactate molecule itself is beneficial. Lactate is a physiologic anion which the body has evolved to utilize as fuel in times of stress. For example, a concentrated solution of sodium lactate has been proven to improve cardiac function (Nalos 2014, Fontaine 2014). Among the various anions included in crystalloid (gluconate, acetate, lactate), lactate is supported by the greatest evidence regarding safety and benefit.
As discussed above, the potassium concentration should be set to an intermediate level of 4 mEq/L (similar to LR).
4. High-normal magnesium concentration
Hospitalized patients usually have hypomagnesemia, which may increase their risk for arrhythmia. One strength of plasmalyte and normosol is the inclusion of magnesium, reducing the need for additional magnesium supplementation. This has been shown to be cost-saving (Smith 2014)(4).
An ideal solution would include a magnesium concentration slightly above the upper limit of normal (1.5 mM, equal to 3.6 mg/dL). Although this might seem a bit high, magnesium levels in this range are entirely safe and potentially desirable. For example, among patients undergoing treatment for atrial fibrillation, a magnesium level of 3.6-4.9 mg/dL may be considered to be the “therapeutic” target range. In practice, this fluid wouldn’t increase the patient’s magnesium level all the way to 3.6 mg/dL, but it would be more effective at increasing the patient’s magnesium to a reasonable level.
Some theoretical and experimental evidence suggests that hyperchloremia may impair renal function.
6. No calcium
LR was designed with a calcium concentration close to physiologic levels. This may be a design flaw for several reasons:
- Including calcium in intravenous fluid causes incompatibility with blood transfusion and certain drugs.
- Giving intravenous calcium usually only causes only a transient increase in the calcium level (5). LR contains a tiny amount of calcium (one liter contains the equivalent of one tenth of a gram of calcium chloride), so it is doubtful that this could have lasting effects on serum calcium level. Even in animal studies comparing large infusions of normal saline vs. LR, no differences in serum calcium were found (Martini 2013).
- Critically ill patients are often hypocalcemic, but this doesn't appear to cause harm. Hypocalcaemia could represent a natural protective mechanism or merely an epiphenomenon of severe illness (e.g. similar to sick euthyroid syndrome). Attempting to raise calcium levels could be harmful, except in cases of iatrogenic hypocalcaemia (e.g. massive transfusion; Aberegg 2016).
My ideal crystalloid
One potential problem with this fluid could be that it is slightly alkalinizing (with a strong ion difference of 34 mEq/L)(6). This is an indirect consequence of being unable to reproduce the anion gap with exogenous albumin (7). However, this shouldn’t be a problem for many reasons:
- My ideal crystalloid (SID 34 mEq/L) is only trivially more alkalinizing than LR (SID 28 mEq/L). For example, the effect of one liter of ideal crystalloid on pH is equivalent to the effect of one liter of LR plus one tenth of an ampule of sodium bicarbonate (8).
- All patients receive at least some normal saline (e.g. used to keep IV lines open and mix drugs). Since normal saline is an acidotic fluid with a strong ion difference of zero, the simultaneous administration of normal saline and ideal crystalloid would likely cancel each other out, with little net impact on the patient’s pH.
- Half-molar sodium lactate (500 mM!) has been shown to be safe and beneficial in patients with heart failure, intracranial hypertension, dengue fever, and post-CABG patients (Nalos 2014, Fontaine 2014). The value of this highly concentrated sodium lactate solution proves that a little extra sodium lactate (e.g., 34 mM vs. 27 mM) could be a good thing (9).
Where is the evidence?
Crystalloid is among the most commonly used drugs in critical care. This creates a paradox: the clinical consequences of crystalloid design out-strip our ability to study them. For example, imagine that adding magnesium reduced the rate of new-onset atrial fibrillation by 0.2% (NNT = 500). This difference would be too small to detect using a RCT (10). However, considering that millions of patients are treated with crystalloid, this difference could nonetheless affect the outcome of thousands of patients (11).
Lacking RCT evidence, the next-best approach is to design a crystalloid based on indirect and physiological evidence. For example, consider the following:
- (A) Normal saline is an acidifying solution with strong ion difference of zero (6). Infusing large volumes of normal saline causes a hyperchloremic metabolic acidosis.
- (B) Clinicians are often uncomfortable if the patient is significantly academic. Sometimes, interventions are performed to maintain the patient's pH (e.g. dialysis, sodium bicarbonate, ventilator adjustments).
It is debatable whether treating acidemia (B) is beneficial. Regardless, it is logically inconsistent to perform a large-volume resuscitation with normal saline (A) and then treat the resulting acidemia (B). There are only two logically coherent approaches to large-volume fluid resuscitation and pH:
- Coherent approach #1: Infuse normal saline, don't monitor pH, and refuse to make any intervention based on the pH or bicarbonate levels.
- Coherent approach #2: Infuse a balanced crystalloid or pH-guided resuscitation fluid, consider intervention to treat the pH and/or bicarbonate levels depending on clinical context.
The best approach is unknown, so it could be reasonable to perform an RCT comparing approach #1 versus approach #2. However, there is little need for an RCT to demonstrate that it is illogical to cause an acidosis (A) and then treat it (B).
A similar argument supports the addition of magnesium to crystalloid. If we are going to measure magnesium levels and replete magnesium, then it is logically consistent to supplement crystalloid with magnesium.
Cost differences are trivial
One argument used to support the use of normal saline is that it is the cheapest. However, the cost differential between LR or normosol versus normal saline is low (<< 1$/liter)(12). Based on the potential to reduce other costs (e.g. magnesium supplementation, occasional avoidance of dialysis)(13), more “expensive” crystalloids are likely to be cost-neutral or cost-saving.
- Currently available fluids were all designed before 1980, prior to a modern understanding of electrolyte and lactate metabolism.
- Saline, LR, and plasmalyte all have some room for improvement.
- Given that crystalloid is administered to millions of patients, even subtle changes could have an effect on some patients.
- It might be time to engineer a new crystalloid, designed to combine the strengths of LR and plasmalyte.
Related posts about electrolytic obsession
- pH-guided resuscitation
- Plasmalyte/Normosol vs. LR
- LR is safe in hyperkalemia
- Understanding lactate & using it to our advantage
- This could also relate to some effect of gluconate (the physiology of which remains quite mysterious). Regardless of the mechanism, this study raises concerns about plasmalyte.
- In practice, this effect is very small for potassium, because absolute differences in potassium concentration are low.
- Of course, once the patient’s serum chemistries are known, it may be wise to choose an aphysiologic solution to intentionally affect the patient’s chemistries (e.g. isotonic bicarbonate to correct a non-anion-gap metabolic acidosis). This concept is discussed further in the section on pH-guided resuscitation.
- As discussed below, the cost differential between different crystalloids is low. In contrast, the material and nursing cost associated with purchasing and infusing separate bags of magnesium is likely to be higher.
- A patient with hypokalemia or hypomagnesemia may often have a total-body deficiency of potassium or magnesium, so it makes some sense to administer potassium or magnesium. In contrast, critically ill patients with hypocalcemia have more complex physiology involving re-distribution of calcium out of the blood. Therefore, treating them with exogenous calcium doesn't seem to work very well – it isn't treating the underlying physiologic problem.
- The strong ion difference (SID) of a fluid is the best way to predict the effect it will have on the body’s pH. A fluid with SID ~24 mEq/L will have little effect on the pH or tend to pull it towards normal. A fluid with SID <24 mEq/L (e.g. normal saline, with a SID of zero) will be acidifying. Alternatively, a fluid with SID >24 mEq/L (e.g. isotonic bicarbonate, with a SID of 150 mEq/L) will be alkalinizing. My ideal crystalloid has a SID of 34 mEq/L, which would tend to be slightly alkalinizing.
- When trying to design a crystalloid that mimics the physiology of blood, the anion gap becomes problematic. Normally the anion gap is largely composed of albumin and phosphate. These cannot be contained in designer crystalloids due to refrigeration and precipitation issues. This makes it difficult to achieve the following three tasks simultaneously: maintain normal osmolality, maintain a normal chloride content, and maintain a normal strong ion difference. Consider, for example, LR and Plasmalyte. LR manages to maintain a normal chloride concentration, at the cost of a slightly elevated strong ion difference (28 mEq/L) and hypotonicity. Plasmalyte gets around this issue by the inclusion of gluconate, but it is questionable whether this “filler” is a physiologically sound approach. My ideal crystalloid approaches the anion gap conundrum by accepting a slightly higher strong ion difference than normal.
- The SID difference between LR and my ideal crystalloid is 6 mEq/L. This is equivalent to 6 ml of 1M sodium bicarbonate, which is nearly one tenth (5 ml) of a standard “amp” of bicarbonate (50 ml of 1 M sodium bicarbonate).
- Incidentally, this data may also suggest that not all alkali are created equal. As discussed above, sodium lactate seems to fare better in clinical trials than sodium bicarbonate or sodium acetate.
- In order to be adequately powered to identify an effect size this low, a study would need to recruit a vast number of patients.
- This is an interesting example of a therapy that could be beneficial despite having a very high NNT (i.e. NNT=500). Generally speaking, if a drug had a NNT of 500, it wouldn't make sense to prescribe it. However, millions of patients need some form of crystalloid. Thus, a minor tweak to crystalloid design could be justified even if it had a very high NNT, because the patients are going to receive crystalloid anyway – it's simply a matter of which formulation.
- Actual prices vary depending on various negotiations between hospitals and manufacturer. In my limited experience, the absolute difference in cost of these crystalloids is very small.
- One of the indications for dialysis is metabolic acidosis. It is well established that balanced crystalloids induce less metabolic acidosis than normal saline. It stands to reason that occasionally, the use of balanced crystalloids (as opposed to normal saline) could tip a patient towards avoidance of dialysis. How frequently this may occur is debatable. However, given the cost of dialysis, even if this occurs very rarely it could justify the increased cost of balanced crystalloids.