Answers are grounded in this paper and cite their sources.
The stress hyperglycemia ratio (SHR) and hemoglobin glycation index (HGI) share HbA1c and glucose inputs. Their mathematical dependence should be distinguished from prognostic association, biological modification and clinical utility. This retrospective MIMIC-IV v3.1 study evaluated adults with documented same-admission heart failure at a 24-hour ICU landmark. SHR used de-duplicated mean glucose during ICU hours [0,24), divided by HbA1c-derived estimated average glucose. Three HGI definitions were compared. Original analyses examined agreement, outcome-free coupling, testing selection, conservative HbA1c-validity restrictions and internal temporal validation. Post-review analyses added paired bootstrap contrasts, formal day-14 interaction, raw-component splines, conditional SHR curves and outcome simulations. The endpoint was database-defined date-level mortality during 28 days after the landmark. Among 14,677 source patients, 3,131 had eligible HbA1c and 2,850 entered the exposure-complete cohort, with 337 coded deaths. HGI-A/B/C correlated with SHR at − 0.605, − 0.791 and − 0.405. The original reference product-term hazard ratio was 0.936 (95% CI 0.881–0.983). Post-review paired HGI-definition contrasts showed no clear differences, whereas testing weighting produced a modest coefficient shift. The late-versus-early product-HR ratio was 1.058 (95% CI 0.959–1.168). Product detection was 6.0% under the no-glycemic-effect simulation and 21.6% and 10.8% under two additive raw-component outcome structures without raw-component interaction. Shared-input algebra is demonstrable, but it does not establish inevitable product-coefficient bias or an independent biological interaction. SHR–HGI joint models should be interpreted as parameterizations of shared glycemic information, with attention to selection and model structure. Not applicable.
No takes yet. Share an insight, caveat, or question.
Zhu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: