The estimated time fraction within range may mask prognostically distinct glycaemic profiles in critical illness: a three-database landmark study.

AIMS: To examine whether out-of-range glucose composition is associated with ICU mortality at a fixed estimated time fraction within range. METHODS: We performed a retrospective 48-h landmark study in MIMIC-IV, eICU and AmsterdamUMCdb. Among adults alive in the ICU at 48 h with adequate glucose measurement coverage, adjusted models assessed post-landmark ICU mortality for a compositional contrast: a 5 percentage point lower time fraction above 180 mg/dL and corresponding higher fraction below 70 mg/dL, holding the fraction within range constant. Time fractions were estimated by linear interpolation of intermittent measurements. RESULTS: The cohorts included 10,732, 17,444 and 4699 patients, with 1079, 1694 and 1015 deaths. The composition contrast was associated with higher mortality, with odds ratios of 1.212, 1.082 and 1.182, respectively. Associations were strongest in days 0-2 and were stronger for recent than remote exposure. Distributional analyses localized the association near and below 54 mg/dL. Shorter intervals reduced precision. CONCLUSIONS: The estimated time fraction within range alone may not fully describe mortality associations. Joint consideration of within-, below- and above-range fractions and their timing may better characterize these associations, a hypothesis that warrants prospective validation.