Early relative glycemic stress trajectories identify a high-risk metabolic phenotype in critically ill patients with atherosclerotic cardiovascular disease: a dual-cohort study.
Early dynamic SHR trajectories identified a common high-risk metabolic phenotype in critically ill ASCVD patients, with the persistently high trajectory independently associated with 28-day all-cause mortality in both derivation and external validation cohorts.
Key Findings
Results
Three distinct early SHR trajectory classes were identified in critically ill ASCVD patients: moderate-stable, low-declining, and persistently high.
Trajectories were derived using group-based trajectory modeling in MIMIC-IV (n=8,835) and externally validated in NWICU (n=1,705).
NWICU patients were assigned to the closest MIMIC-IV-derived trajectory rather than having independent trajectories derived.
The persistently high trajectory was the largest class in both cohorts.
Trajectories were based on SHR values measured across at least three of six 12-hour windows during the first 72 hours after ICU admission.
Results
The persistently high SHR trajectory showed the poorest 28-day survival compared to the other trajectory classes.
In MIMIC-IV, the fully adjusted hazard ratio for Class 3 (persistently high) versus Class 1 (moderate-stable) was 1.37 (95% CI 1.15–1.64).
In NWICU, the corresponding hazard ratio was 2.04 (95% CI 1.39–3.01).
The primary outcome was 28-day all-cause mortality.
Cox models were adjusted for demographic characteristics, ICU type, comorbidities, severity and functional scores, vital signs, laboratory measurements, and early interventions.
Results
The association between persistently high SHR trajectory and mortality was directionally concordant across ICU and in-hospital mortality endpoints.
Findings for ICU mortality and in-hospital mortality were described as 'directionally concordant' with the 28-day mortality results.
This consistency was observed across both the MIMIC-IV derivation cohort and the NWICU validation cohort.
Sensitivity and subgroup analyses were also performed to support robustness of findings.
Results
HbA1c and first-12-hour maximum glucose were the leading predictors of membership in the persistently high SHR trajectory.
These two variables were identified as leading predictors through exploratory machine-learning analyses.
HbA1c reflects the chronic glycemic background used in the SHR denominator, while first-12-hour maximum glucose reflects acute elevation.
The SHR is defined as capturing acute glucose elevation relative to chronic glycemic background as indexed by HbA1c.
Patients required available HbA1c measurements and valid SHR values for inclusion in the study.
Methods
The study included 8,835 patients from MIMIC-IV and 1,705 patients from NWICU, all adults with ASCVD and available HbA1c and SHR data.
This was a retrospective dual-cohort study using MIMIC-IV v3.1 for trajectory development and NWICU v0.1.0 for external validation.
Inclusion criteria required adults with ASCVD, available HbA1c measurements, and valid SHR values in at least three of six 12-hour windows during the first 72 hours of ICU admission.
The dual-cohort design was intended to assess generalizability of trajectory-based risk stratification.
Discussion
The persistently high SHR trajectory reflected acute glycemic deviation from the HbA1c-defined background and was independently associated with short-term mortality.
The stress hyperglycemia ratio (SHR) is designed to capture acute glucose elevation relative to chronic glycemic background.
The persistently high trajectory maintained elevated SHR across the first 72-hour ICU period, distinguishing it from the moderate-stable and low-declining patterns.
The association remained after full adjustment for demographic characteristics, ICU type, comorbidities, severity and functional scores, vital signs, laboratory measurements, and early interventions.
The authors conclude that 'dynamic SHR profiling may support metabolic risk stratification in cardiovascular critical illness.'
What This Means
This research suggests that tracking how a patient's blood sugar relates to their long-term blood sugar control during the first three days in the intensive care unit can identify heart disease patients who are at much higher risk of dying within 28 days. The study used a measure called the stress hyperglycemia ratio (SHR), which compares a patient's current blood sugar to their usual long-term blood sugar level (measured by HbA1c). By analyzing SHR at multiple time points over 72 hours, researchers found three distinct patterns, or 'trajectories': one that stayed moderately stable, one that started low and declined, and one that stayed persistently high. Patients in the persistently high group had significantly worse survival outcomes. This finding held up in two separate patient databases — one used to develop the analysis (over 8,800 patients) and one used to test whether it applied more broadly (over 1,700 patients).
The key insight is that it is not just a single blood sugar reading that matters, but the pattern over time and how it compares to a patient's individual baseline. Patients whose blood sugar remained consistently elevated relative to their chronic baseline were 37% to 104% more likely to die within 28 days compared to those with a moderate-stable pattern, even after accounting for many other health factors. The two strongest signals for predicting which patients would fall into the high-risk group were the HbA1c value itself and the maximum blood glucose in the first 12 hours of ICU admission.
This research suggests that routinely monitoring SHR trajectories — rather than relying on single glucose measurements — could help clinicians identify critically ill heart disease patients who need more intensive metabolic management. The approach could potentially be incorporated into existing ICU monitoring systems, since it uses data (glucose levels and HbA1c) that are already commonly collected. However, as a retrospective observational study, it cannot prove that changing glucose management based on these trajectories would improve outcomes, which would require future clinical trials to confirm.
Wu Y, Wang K. (2026). Early relative glycemic stress trajectories identify a high-risk metabolic phenotype in critically ill patients with atherosclerotic cardiovascular disease: a dual-cohort study.. Frontiers in endocrinology. https://doi.org/10.3389/fendo.2026.1909554