Skip to main content

Foundational Trials

SMART

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Balanced Crystalloids versus Saline in Critically Ill Adults
  • Acronym: SMART (Isotonic Solutions and Major Adverse Renal Events Trial)
  • Year: 2018
  • Journal published in: The New England Journal of Medicine
  • Citation: Semler MW, Self WH, Wanderer JP, et al; SMART Investigators and Pragmatic Critical Care Research Group. Balanced crystalloids versus saline in critically ill adults. N Engl J Med. 2018;378(9):829-839.

Context & Rationale

  • Background
    • 0.9% saline (chloride 154 mmol/L) had been the dominant isotonic crystalloid in ICUs despite supraphysiological chloride exposure.
    • Balanced crystalloids (e.g., lactated Ringer’s; Plasma-Lyte A) more closely match plasma chloride and include buffer anions; biological plausibility suggested lower risk of hyperchloraemic acidosis and renal vasoconstriction.
    • Pre-trial evidence consisted largely of observational studies and small/underpowered randomised comparisons; uncertainty persisted about whether fluid “chloride load” meaningfully affected patient-centred outcomes (kidney failure, survival).
    • Because crystalloids are given to very large numbers of critically ill patients, even small absolute effects could translate into substantial population-level benefit or harm.
  • Research Question/Hypothesis
    • Does a unit-level default strategy of balanced crystalloids (vs 0.9% saline) for isotonic crystalloid therapy in adult ICUs reduce Major Adverse Kidney Events within 30 days (MAKE30: death, new renal-replacement therapy, or persistent renal dysfunction)?
    • Hypothesis: balanced crystalloids would reduce MAKE30 compared with saline when delivered early and pragmatically as routine care.
  • Why This Matters
    • AKI and death are common in critical illness; crystalloid choice is a ubiquitous, system-level intervention.
    • Balanced crystalloids were already available and implementable as an electronic default; SMART tested whether this “small change” had measurable patient-centred benefit.
    • SMART was deliberately paired with a companion emergency department trial (SALT-ED), enabling assessment across acuity settings.

Design & Methods

  • Research Question: In adult ICU patients, does default use of balanced crystalloids (vs 0.9% saline) for isotonic crystalloids reduce MAKE30?
  • Study Type: Pragmatic, cluster-randomised, multiple-crossover trial embedded in routine care; single academic centre; five adult ICUs; unit-level monthly fluid assignment.
  • Population:
    • Setting: five adult ICUs (medical, cardiac, neurological, trauma, surgical) at Vanderbilt University Medical Center.
    • Inclusion: all adults (≥18 years) admitted to participating ICUs during the study period (1 June 2015 to 30 April 2017).
    • Exclusions: none for enrolment; clinician override permitted when a specific fluid was indicated/contraindicated; for kidney components, patients with pre-existing renal-replacement therapy before ICU admission were ineligible to meet “new RRT” and “persistent renal dysfunction” criteria (but remained eligible for mortality).
  • Intervention:
    • Unit-level default to balanced crystalloids (predominantly lactated Ringer’s; Plasma-Lyte A available) for IV isotonic crystalloids in the ICU (bolus and maintenance orders).
    • Electronic order sets and stocking practices aligned to deliver assigned fluid early in critical illness.
    • Non-assigned fluid permitted at clinician discretion (open-label pragmatic delivery).
  • Comparison:
    • Unit-level default to 0.9% saline for IV isotonic crystalloids in the ICU (bolus and maintenance orders).
    • Non-assigned fluid permitted at clinician discretion.
  • Blinding: Unblinded (treating clinicians and units aware of monthly assignment); outcomes were largely objective and EHR-derived.
  • Statistics: Power: 14,000 ICU admissions planned to detect a 1.9% absolute reduction in MAKE30 (15.0% to 13.1%; ≈12% relative reduction) with 90% power at a two-sided 5% alpha; primary analysis was intention-to-treat using regression methods accounting for cluster/multiple-crossover structure and prespecified covariate adjustment.1
  • Follow-Up Period: Primary outcome assessed through hospital discharge or 30 days after ICU admission (whichever occurred first); mortality also reported through 60 days; ICU/ventilator/vasopressor/RRT-free days through day 28.

Key Results

This trial was not stopped early. Planned interim monitoring occurred; enrolment continued to the prespecified sample size.

Outcome Balanced crystalloids 0.9% saline Effect p value / 95% CI Notes
Major adverse kidney event within 30 days (MAKE30) 1139/7942 (14.3%) 1211/7860 (15.4%) Adjusted OR 0.90 95% CI 0.82 to 0.99; P=0.04 Composite: in-hospital death (censored at day 30), new RRT, or final creatinine ≥200% baseline (censored at day 30/discharge).
In-hospital death before day 30 818/7942 (10.3%) 875/7860 (11.1%) Adjusted OR 0.90 95% CI 0.80 to 1.01; P=0.06 Objective component of MAKE30; did not meet conventional statistical significance.
New receipt of renal-replacement therapy 189/7558 (2.5%) 220/7458 (2.9%) Adjusted OR 0.84 95% CI 0.68 to 1.02; P=0.08 Among patients not known to have received RRT prior to ICU admission.
Final creatinine ≥200% of baseline (persistent renal dysfunction criterion) 437/7558 (5.8%) 494/7458 (6.6%) Adjusted OR 0.96 95% CI 0.84 to 1.11; P=0.60 Final inpatient creatinine before discharge (censored at day 30); among patients not known to have received RRT prior to ICU admission.
In-hospital death before ICU discharge 528/7942 (6.6%) 572/7860 (7.3%) Adjusted OR 0.89 95% CI 0.78 to 1.02; P=0.08 Secondary mortality endpoint.
In-hospital death before day 60 928/7942 (11.7%) 975/7860 (12.4%) Adjusted OR 0.92 95% CI 0.83 to 1.02; P=0.13 Secondary mortality endpoint.
Renal-replacement-therapy–free days to day 28 Mean 25.0 ± 8.6 Mean 24.8 ± 8.9 Adjusted OR 1.11 95% CI 1.02 to 1.20; P=0.01 Higher is better (0 assigned if death before day 28); analysed as an ordinal outcome.
Stage 2+ AKI developing after enrolment 807/7558 (10.7%) 858/7458 (11.5%) Adjusted OR 0.91 95% CI 0.82 to 1.01; P=0.09 KDIGO creatinine-based criteria; secondary renal endpoint.
ICU-free days to day 28 Mean 21.8 ± 8.3 Mean 21.7 ± 8.6 Adjusted OR 1.00 95% CI 0.89 to 1.13; P=0.94 No apparent difference in time alive out of ICU.
Ventilator-free days to day 28 Mean 24.2 ± 8.6 Mean 23.9 ± 8.9 Adjusted OR 1.06 95% CI 0.97 to 1.16; P=0.22 No clear difference in ventilation duration at population level.
Vasopressor-free days to day 28 Mean 24.7 ± 8.5 Mean 24.4 ± 8.8 Adjusted OR 1.05 95% CI 0.97 to 1.14; P=0.26 No clear difference in vasopressor duration at population level.
  • Absolute difference for the primary composite was modest (14.3% vs 15.4%), with the composite reaching statistical significance while individual components were directionally consistent but not individually significant.
  • Mechanistic separation aligned with hypothesised chloride effects: hyperchloraemia (>110 mmol/L) occurred less frequently with balanced crystalloids (24.5%) than saline (35.6%), while hyperkalaemia (>5.0 mmol/L) was similar (24.7% vs 25.0%).
  • Pre-specified subgroup: in patients with sepsis or septic shock at ICU admission, MAKE30 was 33.8% (395/1167) with balanced crystalloids vs 38.9% (455/1169) with saline; OR 0.80; 95% CI 0.67 to 0.94; P=0.01 (interaction P=0.06); in traumatic brain injury, MAKE30 was 15.0% (105/698) vs 14.0% (93/665); OR 1.09; 95% CI 0.81 to 1.47; P=0.58 (interaction P=0.24).

Internal Validity

    • Randomisation and allocation: Cluster randomisation at the ICU-month level with repeated crossovers; allocation could not be concealed from clinicians because the intervention was an overt default fluid strategy.
    • Dropout/exclusions: Pragmatic “all-comers” enrolment reduced selection bias; multiple ICU admissions per patient occurred (15,802 ICU admissions among 13,949 unique patients); sensitivity analysis restricted to first ICU admission showed attenuation of the primary effect (reported as not statistically significant).
    • Performance/detection bias: Open-label design introduced potential performance bias; however, the primary outcome used objective endpoints (death, creatinine values, RRT receipt), reducing detection bias; RRT initiation remained clinician-mediated and plausibly susceptible to expectation bias.
    • Protocol adherence: High adherence to assigned crystalloid ordering (assigned fluid accounted for 93.4% of isotonic crystalloid orders during balanced months vs 97.8% during saline months).
    • Separation of the variable of interest: Clear crystalloid separation over time, with marked divergence in assigned fluid exposure (e.g., day 30 saline volume median 0 [IQR 0–77] ml in the balanced group vs 1020 [IQR 0–3500] ml in the saline group; day 30 balanced-crystalloid volume median 1000 [IQR 0–3378] ml vs 0 [IQR 0–0] ml, respectively).
    • Baseline characteristics: Groups were closely matched (examples: median age 58 vs 58 years; male sex 54.4% vs 54.3%; mechanical ventilation at baseline 20.4% vs 21.0%; vasopressors 15.0% vs 15.6%; sepsis/septic shock 14.7% vs 14.9%; median predicted in-hospital mortality 8.7% vs 8.9%).
    • Timing and dose: EHR defaults enabled early delivery; total isotonic crystalloid volumes were moderate in many patients (most medians close to 1 litre over 30 days), which may constrain effect size and accentuate relative importance of high-volume subgroups.
    • Outcome assessment and missingness: Death and RRT were complete; creatinine-based outcomes required baseline and final creatinine estimation; baseline creatinine was missing in a substantial minority and was imputed using prespecified methods, with sensitivity analyses demonstrating some dependence of effect estimates on creatinine handling.
    • Statistical rigour: Analyses accounted for clustering/crossover and included prespecified covariate adjustment; the modest P value for the primary endpoint (0.04) increases interpretability sensitivity to multiplicity/interim-monitoring conventions, though the prespecified plan included interim monitoring.

Conclusion on Internal Validity: Overall, internal validity appears moderate to strong: pragmatic enrolment and objective endpoints support robustness, and protocol adherence achieved clear separation; key limitations are open-label delivery, clinician-mediated RRT initiation, single-centre design, and sensitivity of creatinine-derived components to baseline estimation.

External Validity

    • Population representativeness: Broad adult ICU case-mix across five specialised units, minimal exclusions, and pragmatic delivery make findings relevant to many mixed medical–surgical ICUs.
    • Single-centre context: Results may depend on local practice patterns (baseline fluid choice, early AKI detection, RRT thresholds, ED/operating theatre alignment with ICU defaults), potentially affecting transportability.
    • Applicability to specific subpopulations: Neurocritical care and traumatic brain injury populations may require tailored fluid strategies (osmolarity/ICP considerations), and later evidence has raised fluid-specific concerns in traumatic brain injury in other settings.
    • Health-system implementation: The intervention is highly implementable where EHR defaults and supply chain can enforce unit-level fluid assignment; in resource-limited settings, availability and cost constraints may differ.

Conclusion on External Validity: Generalisability is good for adult mixed ICUs in high-resource systems but is limited by the single-centre nature and potential dependence on local RRT practices and pre-ICU fluid exposure.

Strengths & Limitations

  • Strengths: Very large pragmatic ICU trial; minimal exclusion and “default strategy” design reduced selection bias; high protocol adherence with strong separation; objective patient-centred composite endpoint; rapid translation potential via EHR defaults; paired programme with ED companion trial enabled broader inference.
  • Limitations: Unblinded and single-centre; cluster-multiple-crossover design can introduce time-varying confounding and contamination across settings; primary effect size modest and driven by a composite with components that were not individually significant; RRT initiation susceptible to clinician discretion; creatinine-based components depend on baseline/last creatinine measurement and imputation strategies; balanced crystalloids comprised more than one formulation without head-to-head randomisation.

Interpretation & Why It Matters

  • Clinical implication
    Using balanced crystalloids as the ICU default produced a modest reduction in MAKE30 compared with saline, supporting balanced crystalloids as a reasonable “default” isotonic crystalloid for many critically ill adults (especially where kidney injury prevention is prioritised).
  • Mechanistic coherence
    Lower rates of hyperchloraemia with balanced crystalloids strengthened biological plausibility that chloride load is a modifiable exposure; however, the translation into hard endpoints was small and context-dependent.
  • Systems-level lesson
    SMART demonstrated that EHR-embedded “default” strategies can be used to execute large, low-friction randomised trials and can serve as a bridge from comparative effectiveness evidence to practice change.

Controversies & Subsequent Evidence

    • Editorial framing (SMART + SALT-ED): The accompanying NEJM editorial interpreted SMART alongside the companion ED trial (SALT-ED), emphasising patient-centred outcomes and the pragmatic value of system-level fluid defaults while highlighting that effect sizes were small and endpoint choice matters.23
    • Endpoint interpretation: MAKE30 is clinically meaningful but composite; the primary signal emerged despite non-significance of individual components, raising interpretability questions typical of composite endpoints where components differ in frequency and clinical weight.
    • Single-centre pragmatism vs reproducibility: The trial’s strength (system default with minimal exclusions) is also a limitation (unblinded, centre-specific pathways for RRT and fluid delivery), increasing the importance of replication.
    • Subgroup signal (sepsis): The larger apparent benefit in sepsis/septic shock (MAKE30 OR 0.80; 95% CI 0.67 to 0.94; interaction P=0.06) is biologically plausible (high fluid volumes, acid-base derangements), but remains hypothesis-generating given multiplicity and modest interaction strength.
    • Later large ICU RCTs: Two subsequent multicentre ICU trials did not show overall mortality benefit with balanced crystalloids: BaSICS (JAMA 2021) and PLUS (NEJM 2022), both reporting neutral primary mortality outcomes and raising the possibility that any benefit is smaller, context-specific, or concentrated in particular subgroups/strategies rather than universally present.45
    • Hospital-wide evidence: The FLUID hospital-wide crossover trial (NEJM 2025; ~159,000 patients) found no clear difference in its primary composite outcome of death or hospital readmission within 90 days (20.3% lactated Ringer’s vs 21.4% saline; adjusted difference −1.1 percentage points; 95% CI −2.4 to 0.2), suggesting that if a benefit exists across broad hospital populations it is likely small and sensitive to context and implementation.6
    • Meta-analytic synthesis: A systematic review/meta-analysis (NEJM Evidence 2022) estimated a small mortality benefit signal for balanced crystalloids versus saline in critical illness (e.g., mortality RR 0.96; 95% credible interval 0.93 to 0.99; posterior probability of reduced mortality ~89.5%), but the absolute effect remained small and heterogeneity by population (e.g., traumatic brain injury) remained an important concern.7
    • Guideline incorporation: Surviving Sepsis Campaign 2021 suggested balanced crystalloids over saline for initial resuscitation in sepsis/septic shock, reflecting the aggregate of SMART/SALT-ED and subsequent evidence.8
    • Updated critical care fluid guidance: ESICM’s 2024 clinical practice guideline on fluid therapy (Part 1: choice of resuscitation fluids) synthesised the post-SMART evidence base (including later large RCTs) and provides an updated framework for selecting resuscitation crystalloids in adult critical illness.9

Summary

  • Pragmatic ICU-default strategy trial comparing balanced crystalloids with 0.9% saline across 15,802 ICU admissions.
  • Primary endpoint (MAKE30) favoured balanced crystalloids: 14.3% vs 15.4%; adjusted OR 0.90; 95% CI 0.82 to 0.99; P=0.04.
  • Individual components (death by day 30; new RRT; persistent renal dysfunction criterion) were directionally consistent but not individually significant.
  • Clear physiological separation was achieved (less hyperchloraemia with balanced crystalloids), with no signal of increased hyperkalaemia.
  • Later multicentre ICU trials and meta-analyses suggest any benefit is small and context-dependent; guidelines increasingly favour balanced crystalloids in sepsis while maintaining nuance for specific populations.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • The NEJM editorial discussed SMART alongside the companion emergency department trial (SALT-ED); the combined programme accelerated uptake of balanced crystalloids as a “default” strategy in many institutions.
  • Balanced crystalloids in SMART were predominantly lactated Ringer’s, with a smaller proportion of Plasma-Lyte A; differential effects between balanced formulations were not directly tested.

Overall Takeaway

SMART was a landmark pragmatic ICU trial showing that changing the default crystalloid from 0.9% saline to balanced solutions yields a small but measurable reduction in a patient-centred kidney composite (MAKE30). Its influence lies as much in the demonstration of scalable, EHR-embedded randomised comparative effectiveness methods as in the clinical effect size, and it catalysed a decade of larger multicentre trials, meta-analyses, and guideline updates refining when balanced crystalloids should (and should not) be the default.

Overall Summary

  • Large pragmatic ICU-default trial: balanced crystalloids modestly reduced MAKE30 versus saline.
  • Physiological separation (less hyperchloraemia) supported biological plausibility; hard endpoints moved only slightly.
  • Subsequent large trials and guidelines suggest benefit, if present, is small and context-dependent (not “one-size-fits-all”).

Bibliography