Balanced Crystalloids or 0.9% Saline in Sepsis. Beyond Reasonable Doubt?

Intravenous fluid therapy with crystalloid solutions is one of the most common interventions for patients with sepsis. Both 0.9% saline and balanced crystalloids are widely used (1). However, with respect to mortality risk, the comparative effectiveness of these fluids is uncertain (2).

In this issue of the Journal, Brown and colleagues (pp. 1487–1495) report a post hoc analysis of SMART (Isotonic Solutions and Major Adverse Renal Events Trial) (3). SMART was a single-center, open-label, cluster-randomized, multiple-crossover trial (4). A total of 15,802 patients were enrolled in five ICUs at Vanderbilt University Medical Center in under 2 years. This remarkable feat was possible because of the study’s novel design. Random assignment to balanced crystalloids or 0.9% saline occurred at the level of the ICU, rather than at the level of the individual patient, and each ICU “crossed over” to use each fluid multiple times over the duration of the study. All patients who were admitted to an ICU during the study were included in the study by default. All data were obtained from the electronic health record, and a waiver of consent was granted. This novel methodology represents a major breakthrough for comparative effectiveness research in critical care and has resulted in a tremendously useful dataset that can now be used for hypothesis-driven exploratory analyses. Here the authors report one such analysis evaluating 30-day in-hospital mortality among the 1,641 trial patients who were admitted to the medical ICU with sepsis. Their hypothesis was that balanced crystalloids would reduce in-hospital mortality at day 30.

The primary finding of the current analysis was that 217 patients (26.3%) in the balanced crystalloids group died in-hospital within 30 days compared with 255 patients (31.2%) in the saline group (adjusted odds ratio, 0.74; 95% confidence interval, 0.59–0.93; P = 0.01) (3). Based on the observed point estimate of treatment effect, this corresponds to a number needed to treat with balanced crystalloids instead of saline to prevent one death of just over 20.

Although the investigators’ findings were robust in a number of sensitivity analyses, as they themselves have highlighted, caution is required in interpreting these findings. One methodological issue of potential concern is that the comparison between treatment groups undertaken was not truly randomized because patients were categorized as having sepsis based on International Statistical Classification of Diseases and Related Health Problems, 10th Revision codes. These codes were determined after randomization had already occurred. Although it seems unlikely that assignment to balanced crystalloids or 0.9% saline affected such coding, it might have, and in doing so, could have introduced bias. According to the findings of previous research, the mechanism of harm from 0.9% saline was hypothesized to be via hyperchloremic metabolic acidosis and renal injury (5, 6). However, the mortality effect in the current analysis was not altered by either the baseline chloride or bicarbonate concentrations, and there was no apparent effect of study treatment on serum creatinine. Although these findings led the authors to conclude that the observed mortality effect might be explained by the hemodynamic effects of the fluids used, an alternative explanation is that the findings in relation to day 30 in-hospital mortality represent a false positive. In general, the volumes of fluids that patients received were very small. In both groups, the median volume of study fluid received in the emergency department was 1,000 ml (interquartile range [IQR], 0–2,000 ml). Between ICU admission and Day 30, patients allocated to 0.9% saline received a median of 2,000 ml (IQR, 500–4,830 ml) of 0.9% saline in total, and patients allocated to balanced crystalloids received a median of 1,500 ml (IQR, 0–4,000 ml) in total. Given the small volumes of fluid administered, one can only conclude either that 0.9% saline is extraordinarily toxic to patients with sepsis or that the mortality effect observed is a result of the play of chance. In favor of the latter explanation, it is notable that the 95% confidence intervals for the odds ratio for in-hospital mortality at Day 60 cross one (i.e., encompass no effect). Moreover, as there was no statistically significant heterogeneity of treatment effect for patients with and without sepsis either in the original SMART study (4) or in the current analysis of medical ICU patients (6), the available data do not support the conclusion that patients with sepsis respond differently to other patients when given 0.9% saline as compared with balanced crystalloids.

Based on these considerations, it is clear that the findings of this post hoc analysis do not constitute proof beyond reasonable doubt that balanced crystalloids reduce in-hospital mortality compared with 0.9% saline in patients with sepsis. Because millions of patients around the world with sepsis receive intravenous crystalloid fluid therapy every year, such proof is desirable to inform healthcare decision-making on a global scale. Although randomized controlled trials that are being conducted in Australia and New Zealand (7) and in Brazil (8), respectively, may provide such proof, despite the caveats outlined here, the current data are already actionable by clinicians. This analysis includes data from 1,641 trial participants with sepsis, and the treatment groups appear very similar at baseline. Differences in outcomes observed could well be a result of differences in exposure to the fluids studied. Moreover, for trials that evaluate the comparative effectiveness of standard treatments, my view is that clinicians should not wait for proof beyond a reasonable doubt before they implement trial findings. Instead, the appropriate burden of proof to consider should be based on the balance of probabilities. Logically, given that in many countries both 0.9% saline and balanced crystalloids are available and widely used, clinicians should ask themselves, “On the basis of all of the available evidence (and considering the costs), is it more probable that 0.9% saline or a balanced crystalloid is the best choice for this patient?”

On the balance of probabilities, the data presented by Brown and colleagues (3) suggest that balanced crystalloids are likely to be the best choice. Although these data do not represent proof beyond a reasonable doubt, they add to a growing body of evidence suggesting that balanced crystalloid fluids are preferred to 0.9% saline for intravenous fluid therapy (4, 9). Although the results of ongoing trials are awaited, clinicians can reasonably use balanced crystalloids in preference to 0.9% saline. Doing so might well reduce the risk for death in critically ill patients with sepsis and potentially in other patients as well.