Objective: To evaluate whether low plasma glutamine (PG) is related to severity of illness, and actual and predicted hospital mortality. Design: Prospective cohort study. Setting: 18-bed closed format general intensive care unit (ICU) of a teaching hospital. Patients: Cohort of 80 seriously ill patients non-electively admitted to the ICU. Interventions: Blood sampling for the determination of PG at ICU admission. Measurements and results: Severity of illness and predicted mortality were calculated using the locally validated APACHE II, SAPS II, and MPM II 0 and 24 systems. Illness scores, and actual and predicted hospital mortality were compared between patients with total PG < 0.420 mmol/l ("low PG") and patients with PG≥0.420 mmol/l. Mean total PG was 0.523 mmol/l, range 0.220-1.780 mmol/l. Low PG (n = 25) was associated with higher age (P = 0.03), shock as primary diagnosis, and higher actual hospital mortality (60% vs 29%, P = 0.01). Normal to high PG was associated with high plasma creatine phosphokinase (P = 0.007). There was a nonsignificant trend towards higher severity of illness scores and predicted mortality rates in the low PG group. The presence of low PG significantly improved mortality prediction when added as a factor to the APACHE II predicted mortality rate (P = 0.02). Conclusions: Low PG at acute ICU admission is related to higher age, shock as primary diagnosis, and higher hospital mortality. Low PG represents a risk of poor outcome, not fully reflected in the presently used mortality prediction system.