Prior exposure to hyperglycaemia attenuates the relationship between glycaemic variability during critical illness and mortality.

OBJECTIVE Our primary objective was to determine the impact of prior exposure to hyperglycaemia on the association between glycaemic variability during critical illness and mortality. Our secondary objectives included evaluating the relationships between prior hyperglycaemia and hyperglycaemia or hypoglycaemia during critical illness and mortality. DESIGN AND PARTICIPANTS A single-centre, retrospective, observational study in a tertiary intensive care unit. Patients admitted to the ICU between 1 September 2011 and 30 June 2015, with diabetes recorded using ICD-10-AM coding or a glycated haemoglobin (HbA1c) level of ≥ 6.5%, were stratified by prior hyperglycaemic level (HbA1c < 6.5%, 6.5%-8.5%, or > 8.5%). MAIN OUTCOME MEASURES Glycaemic variability was assessed as the blood glucose coefficient of variation during the patient's stay in the ICU. Multivariate logistic regression and marginal predictive plots were used to assess the impact of prior hyperglycaemia on the association between glycaemic metrics and mortality. RESULTS We studied 1569 patients with diabetes (HbA1c < 6.5%, n = 495; HbA1c 6.5%-8.5%, n = 731; and HbA1c > 8.5%, n = 343). Glycaemic variability was strongly associated with hospital mortality (P = 0.001), but this asso ciation showed a significant interaction with prior hyperglycaemia (P = 0.011), such that for patients with HbA1c > 8.5%, increasing glycaemic variability was not associated with increased mortality. Acute hyperglycaemia was strongly associated with mortality (P < 0.0001) and also showed a significant interaction with prior hyperglycaemia (P = 0.001), such that for patients with HbA1c > 8.5%, acute hyperglycaemia was not associated with mortality. Hypoglycaemia was also associated with mortality (P < 0.0001), but prior exposure to hyperglycaemia had a lesser effect on this relationship. CONCLUSION Prior exposure to hyperglycaemia attenuates the association between glycaemic variability and mortality in critically ill patients with diabetes.