Using 0.9% saline has been undergoing scrutiny recently, with detractors claiming that high chloride content leads to acidosis and kidney injury and clinical evidence supporting the absence of these deleterious effects. (JAMA 2015;314:1701; http://bit.ly/2N6JOtR.) Two recent massive single-center trials examining the utility of balanced solutions compared with normal saline, however, provide little guidance for clinicians about which tact to take.
The first, the SALT-ED trial, was a pragmatic, nonblinded, multiple-crossover trial that enrolled adults who received at least 500 mL of intravenous isotonic crystalloids in the ED and subsequently were hospitalized outside of an ICU. (N Engl J Med 2018;378:819; http://bit.ly/2zuvObg.) All patients in the trial that same month were assigned to the same fluid, either balanced crystalloids or saline. The authors enrolled 13,347 patients during the 16-month trial period, assigning 6,708 (50.3%) to receive balanced crystalloids and 6,639 (49.7%) to saline. Patients generally received between one and two liters during their ED stay.
No difference was reported in the primary outcome—hospital-free days—but the researchers did note that patients in the balanced crystalloids group experienced a 0.9 percent lower incidence of major adverse kidney events within 30 days (MAKE-30, defined as the composite of death, new receipt of renal-replacement therapy, or persistent renal dysfunction, 4.7% vs. 5.6%).
The second trial, SMART, examined the use of balanced crystalloid solutions only in patients admitted to the ICU. (N Engl J Med 2018;378:829; http://bit.ly/2NMhwpK.) This pragmatic, nonblinded, cluster-randomized, multiple-crossover trial was conducted in five ICUs at a single medical center. Participating ICUs were randomly assigned to use saline during even-numbered months and balanced crystalloids during odd, or vice versa.
Similar to the previous trial, the authors noted a statistically significant difference in MAKE-30: 14.3 percent v. 15.4 percent in the balanced solution and saline groups, respectively. Like the SALT-ED trial, this 1.1 percent absolute difference reached statistical significance with a p value of 0.04.
It is important to note that both trials found success in a composite outcome, which offers a number of benefits. It allows for the combination of multiple event types into a single outcome, increasing the effect size and potentially augmenting the efficiency and statistical power of any sample size, and it provides an additional benefit of combining multiple endpoints into one primary outcome without violating the singular nature of the primary endpoint, in this case, new renal-replacement therapy or persistent renal dysfunction.
Composite endpoints augment statistical precision, but they come at the cost of added clinical uncertainty. Each individual component of a composite outcome is weighted the same, placing equal value on events of varying degrees of clinical importance. In this case, there is a dramatic difference in the clinical importance of a rise in creatinine of greater than 200 percent from baseline, a purely disease-oriented outcome, to death, an outcome of true patient-centered relevance.
The SALT-ED trial authors found a 0.9 percent decrease in the rate of MAKE-30 (p value 0.01) in ED patients who received a balanced crystalloid solution but no difference in 30-day mortality between the patients in the balanced solution and saline groups (1.4% and 1.5%, respectively). Nor was any difference seen in the number of patients who required new renal replacement therapy (0.3% vs 0.5%). In fact, the entire difference was powered by a 0.7 percent increase in the rate of patients experiencing a 200 percent increase in the creatinine above baseline at 30 days. The study enrolled 13,347 patients, and the only difference identified was a minute decrease in a clinically questionable outcome.
The argument for using a composite outcome is that the events being measured are rare, sometimes prohibitively so, requiring an impossibly large sample to identify the true effect size. Assembling an assortment of outcomes all representing a shared underlying physiological benefit provide the ability to identify small benefits of the treatment in question that would otherwise go unnoticed. But if the composite outcome does not represent a unified physiological process, it is in danger of piecing together small random variations in effect size, augmenting statistical chance rather than identifying a clinical reality.
The SMART trial reported a statistically significant decrease in MAKE-30 events in patients who received a balanced crystalloid solution. Similar to the SALT-ED trial, none of the individual components that made up the MAKE-30 outcome was independently significant. A p value less than 0.05 was achieved only when added together. It also had a 0.8 percent absolute difference in 30-day mortality, a 0.4 percent difference in new renal replacement therapy, and only a 0.2 percent difference in the rate of persistent kidney injury.
One cannot help noticing the lack of difference in the rates of kidney injury between the two groups. No difference was seen in renal replacement-free days, the incidence of stage 2 or higher AKI, highest creatinine level, or change in creatinine level above baseline between the two groups. If the MAKE-30 criteria were intended to identify the clinically significant deleterious effects of saline on renal function, then one has to ask the obvious question, where is this renal dysfunction? If no difference is seen in the rates of AKI between the two groups, can we confidently state that combining these three metrics into one composite is an appropriate clinical measure?
The final answer, of course, is unclear. The SMART trial offered us statistically significant results, but it is important to remember how they reached this answer: By enrolling 15,802 patients in a nonblinded, pseudo-randomized study using a composite outcome in which the individual components demonstrated minute differences of questionable clinical relevance.
The real question to ask is, do these results represent a true clinical effect or just statistical noise stacked on top of random chance, piled on happenstance, that when assembled one on top of the other offers the false sense of clinical import?Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.