Clinical Implications of Low Anion Gap in Critically Ill Patients: Insights from the BOLD Multi-Centre Blood Gas Dataset.

Objectives: Low and subnormal anion gap (AG) values are frequently overlooked in clinical practice despite potential diagnostic and prognostic implications. We aimed to characterise the distribution of AG and albumin-corrected anion gap (CAG) in a large multicentre cohort of critically ill patients, quantify the prevalence of low and subnormal values and examine AG behaviour in the presence of clinically significant lactic acidosis. Materials and methods: We performed a retrospective secondary analysis of the publicly available Blood Gas and Oximetry Linked Dataset (BOLD), comprising de-identified records from MIMIC-III, MIMIC-IV and eICU-CRD databases. Anion gap and CAG were calculated, with CAG being adjusted for albumin. Low AG was defined as <6 mmol/L, subnormal as 6-10 mmol/L and elevated as >16 mmol/L. Clinically significant lactic acidosis was defined as lactate >5 mmol/L. Mortality differences were assessed using Pearson's chi-square test. Results: Among 37,763 admissions with available AG data, AG was less than 6 mmol/L in 1.6% of patients and <10 mmol/L in 15.3% of subjects. After albumin correction, CAG was less than 6 mmol/L in 0.5% of patients and <10 mmol/L in 5.8% of subjects. In the subgroup with lactate >5 mmol/L, 19.2% of patients did not have an elevated AG and 8.2% did not exhibit an increased CAG. Among patients with normal or low CAG values, in multivariable logistic regression adjusting for age, pH, albumin, SOFA score and comorbidity index, clinically significant lactic acidosis remained independently associated with mortality [adjusted odds ratio (OR) 3.05, 95% confidence interval (CI) 2.04-4.56]. Conclusions: Low and subnormal AG values are relatively common in critically ill patients even after albumin correction and the causes remain unclear. A substantial proportion of patients with clinically significant lactic acidosis do not exhibit an elevated AG. Clinicians should exercise caution when using AG to categorise metabolic disorders, as high-anion-gap physiology may be masked by an underlying low or subnormal anion gap.