Serial Lactate-to-Albumin Ratio Adds Signal, but Not a Clear New Hierarchy, in Critical Illness
In critically ill adults, a rising lactate-to-albumin ratio has been treated as a shorthand for severe physiologic stress: lactate reflects impaired perfusion and anaerobic metabolism, while albumin falls with inflammation, capillary leak, malnutrition, and hepatic dysfunction. This multicohort study tested whether early longitudinal patterns of lactate-to-albumin ratio could sharpen mortality prediction beyond a single measurement. The answer was mixed. Higher mean LAR tracked with death in the derivation cohort, yet the added prognostic value over lactate alone remained modest, and the trajectory story changed depending on how time was anchored.
The work used two large critical care databases and aligned exposure windows carefully to avoid mixing early physiologic instability with later outcome follow-up. In MIMIC-IV, LAR was assessed during the first 0 to less than 96 hours, and mortality follow-up began at 96 hours. Among 1,456 landmark-eligible patients, a higher mean LAR was associated with both 30-day and 365-day mortality. The adjusted risk ratio per 1 standard deviation increase was 1.22 for 30-day mortality and 1.09 for 365-day mortality. Those associations, however, were similar for mean lactate, which immediately weakens any claim that the combined ratio is inherently superior to its metabolic component.
A deeper biological reading still makes sense. Lactate rises when oxygen delivery, mitochondrial function, or clearance is compromised; albumin falls in parallel with systemic inflammation and physiologic reserve. Combining the two can therefore capture a patient who is both overproducing stress metabolites and underrepresented by a protein marker of chronic and acute illness burden. In an ICU bed, that combination may identify patients whose hemodynamics, inflammatory state, and nutritional reserve are all drifting in the wrong direction at once. Yet a useful biomarker must do more than mirror severity. It has to change decisions or outperform the simpler marker already available at the bedside.
Trajectory Classes Shifted with Time, Exposing a Fragile Pattern Structure
The study moved beyond a single mean value and built a descriptive three-class trajectory model. In MIMIC-IV, the classes were stable-low, moderate-declining, and high-increasing. Neither non-reference trajectory remained associated with increased mortality after adjustment. Then the pattern structure shifted under alternative temporal definitions, a reminder that longitudinal phenotypes can be highly sensitive to where the clock starts and stops.
This is not a minor statistical quirk. ICU biomarker trajectories are shaped by resuscitation, source control, vasopressor titration, transfusion, fluid balance, and the timing of laboratory draws. When one moves the temporal window, the cohort may stop describing the same physiologic state. A stable-low pattern may simply reflect rapid stabilization after early shock, while a high-increasing pattern may capture delayed deterioration, persistent hypoperfusion, or worsening capillary leak. But if the class labels do not survive a change in timing, confidence in a universal trajectory phenotype falls quickly.
The independent eICU cohort reinforced that concern. With _n_ = 1,181, trajectory estimates differed from those in MIMIC-IV, although a stable-high class was associated with higher hospital and ICU mortality. Even there, the signal was not a clean endorsement of a single transferable phenotype across systems. Different patient mixes, measurement practices, and ICU workflows likely shaped the pattern space, and the ratio itself may be reflecting local data capture as much as biology.
Lactate Carried Most of the Prognostic Weight
A blunt finding emerged from the secondary prediction analyses: mean LAR offered only modest incremental performance, and trajectory class alone contributed little. That matters at the bedside because lactate is already embedded in critical care practice, while albumin can be influenced by acute dilution, liver synthetic function, and chronic illness in ways that complicate interpretation. If the ratio does not reliably outperform lactate, then the extra calculation may not justify replacing a simpler marker in everyday prognostic triage.
There is still a practical physiologic appeal to the ratio. Hypoalbuminemia may amplify the meaning of a given lactate level by signaling lower reserve, more inflammation, and less buffering capacity against critical illness. A patient with moderate lactate elevation and very low albumin may be very different from one with the same lactate and preserved albumin. But this study suggests that distinction is not yet stable enough to become a general rule for risk stratification, especially when longitudinal structure changes with the analytic framework.
The selected population also deserves attention. These were critically ill adults with sufficient serial measurements, which means the cohort was not an unselected ICU population. Patients without repeated labs, or those who stabilized quickly enough to escape repeated testing, were not the main focus. Prognostic markers often look stronger in cohorts enriched for serial data because clinicians are already sampling the sickest patients most often. That selection can make a biomarker appear more informative than it would be in routine practice.
What This Means for ICU Risk Stratification
For now, early serial lactate-to-albumin ratio should be viewed as a potentially informative marker of risk, not as a replacement for lactate or a definitive longitudinal phenotype. The ratio may help describe a patient’s physiologic burden, especially when inflammation and hypoperfusion coexist, but this study does not establish clinical utility. Its strongest message is restraint: the biomarker signal is real, yet it is not robust enough to claim superiority over lactate or to support a universal trajectory scheme across databases.
That leaves a clear research problem rather than a finished clinical tool. Future work will need prospectively timed sampling, standardized landmark definitions, and validation in cohorts where albumin kinetics can be interpreted alongside resuscitation milestones and organ support. Without that, the ratio remains a plausible but incomplete marker of danger in critical illness, one that can describe risk but still struggles to outperform the simplest number already on the chart.