Objective: To study the effect of stress hyperlactatemia on the association between stress hyperglycemia and mortality.
Design: Retrospective cross-sectional observation study.
Setting: Three ICUs using arterial blood gases with simultaneous glucose and lactate measurements during ICU stay.
Patients: Cohort of 7,925 consecutive critically ill patients.
Measurements and main results: We analyzed 152,349 simultaneous measurements of glucose and lactate. We performed multivariable analysis to study the association of different metrics of glucose and lactate with hospital mortality. On day 1, first (p = 0.013), highest (p = 0.001), mean (p = 0.019), and time-weighted mean (p = 0.010) glucose levels were associated with increased mortality. A similar, but stronger, association was seen for corresponding lactate metrics (p < 0.0001 for all). However, once glucose and lactate metrics were entered into the multivariable logistic regression model simultaneously, all measures of glycemia ceased to be significantly associated with hospital mortality regardless of the metrics being used (first, highest, mean, time-weighed; p > 0.05 for all), whereas all lactate metrics remained associated with mortality (p < 0.0001 for all). In patients with at least one episode of moderate hypoglycemia (glucose ≤ 3.9 mmol/L), glucose metrics were not associated with mortality when studied separately (p > 0.05 for all), whereas lactate was (p < 0.05 for all), but when incorporated into a model simultaneously, highest glucose on day 1 was associated with mortality (p< 0.05), but not other glucose metrics (p > 0.05), whereas all lactate metrics remained associated with mortality (p < 0.05 for all).
Conclusions: Stress hyperlactatemia modifies the relationship between hyperglycemia and mortality. There is no independent association between hyperglycemia and mortality once lactate levels are considered.