Prognostic Association and Predictive Value of the Stress Hyperglycemia Ratio for All‑Cause Mortality in Elderly Hip Fracture Patients with Perioperative Atrial Fibrillation

Technical Summary: Prognostic Association and Predictive Value of the Stress Hyperglycemia Ratio in Elderly Hip Fracture Patients with Perioperative Atrial Fibrillation

Problem Statement
Elderly patients with hip fractures face high mortality rates, a risk further exacerbated by the presence of perioperative atrial fibrillation (AF). While the Stress Hyperglycemia Ratio (SHR)—defined as admission fasting blood glucose divided by an estimated average glucose derived from HbA1c—has been linked to adverse outcomes in cardiovascular and critical care settings, its specific prognostic significance in the intersection of geriatric orthopedics and perioperative AF remains undefined. Furthermore, the nature of the association (linear vs. nonlinear) between SHR and long-term mortality in this specific cohort is unclear, and the incremental predictive value of SHR beyond standard clinical variables has not been established.

Methodology
This retrospective cohort study analyzed 340 patients aged ≥65 years admitted to the Third Hospital of Hebei Medical University between January 2019 and December 2022, who presented with hip fractures and perioperative AF. The primary outcome was all-cause mortality within 36 months post-discharge.

The analytical approach employed a multi-stage statistical framework to address confounding and nonlinearity:

1. Confounding Adjustment: Generalized Propensity Score Weighting (GPSW) was utilized to balance covariates across the continuous distribution of SHR, mitigating systematic differences in baseline characteristics (e.g., albumin levels).
2. Nonlinearity Visualization: Restricted Cubic Spline (RCS) curves with four knots were constructed to visualize the dose-response relationship between SHR and mortality, both in unadjusted and GPSW-adjusted models.
3. Variable Selection: To handle the nonlinear nature of SHR, the RCS basis functions were grouped and entered into a Group Lasso-Cox regression. This method performs group-wise penalization, selecting the entire spline basis group if it contributes to the model, thereby avoiding the limitations of traditional Lasso in capturing nonlinear effects.
4. Model Validation and Comparison:
   • Internal Validation: Performed via bootstrap resampling (500 replicates) to assess model stability and calculate optimism-corrected C-indices.
   • Predictive Performance: Time-dependent Receiver Operating Characteristic (td-ROC) curves were generated at 12 and 24 months.
   • Incremental Value: A reduced model excluding the SHR spline basis group was compared against the full model using C-indices and Time-dependent Decision Curve Analysis (td-DCA) to evaluate net clinical benefit.

Key Results

• Nonlinear Association: RCS analysis revealed a significant J-shaped nonlinear association between SHR and all-cause mortality. In the unweighted model, the hazard ratio (HR) equaled 1 at SHR cutoffs of 0.74 and 1.21; these values remained robust (0.71 and 1.19) after GPSW adjustment. This suggests that both low and high extremes of SHR are associated with increased mortality risk compared to a moderate range.
• Predictor Selection: The Group Lasso-Cox model identified nine predictor groups: the SHR spline basis group, age, coronary artery disease (CAD), chronic heart failure (CHF), acute myocardial infarction (AMI), ischemic stroke, B-type natriuretic peptide (BNP), left ventricular end-diastolic diameter (LVEDD), and left ventricular ejection fraction (LVEF).
• Model Performance:
  • The full model (including SHR) achieved an original C-index of 0.7688, compared to 0.7229 for the model excluding SHR.
  • After bootstrap correction, the SHR-inclusive model maintained a superior C-index (0.7554) over the reduced model (0.7118).
  • Time-dependent AUCs for the SHR-inclusive model were 0.720 at 12 months and 0.719 at 24 months.
• Clinical Utility: Decision curve analysis confirmed that the model incorporating SHR provided a superior net clinical benefit across a broad range of threshold probabilities at both 12 and 24 months compared to the model without SHR.

Key Contributions

1. First Investigation in Specific Cohort: This study represents the first retrospective analysis exploring the relationship between SHR levels and long-term all-cause mortality specifically in elderly hip fracture patients with perioperative AF.
2. Methodological Rigor: The study advances the statistical handling of SHR in survival analysis by combining GPSW for confounding control with Group Lasso-Cox for variable selection. This approach effectively captures the J-shaped nonlinear effect of SHR without violating linearity assumptions or overfitting, a common challenge in small-sample survival studies.
3. Incremental Predictive Value: The study demonstrates that SHR is not merely a correlate but an independent predictor that significantly enhances the discriminative ability and clinical net benefit of prognostic models for this high-risk population.

Significance and Claims
The authors claim that SHR exhibits a J-shaped relationship with long-term mortality in this specific population, indicating that maintaining SHR within a moderate range may be associated with improved survival. The study concludes that SHR provides incremental prognostic value, suggesting its inclusion in risk stratification models for elderly hip fracture patients with AF.

In terms of clinical practice, the authors propose that SHR should be recognized as an independent risk factor. They suggest that for elderly hip fracture patients with AF, even those without a history of diabetes, blood glucose levels should be closely monitored. Timely intervention to manage stress hyperglycemia, alongside individualized management considering comorbidities, may improve rehabilitation and prognosis. However, the authors modestly note limitations, including the single-center retrospective design, the relatively small sample size due to the low prevalence of AF in this context, and the lack of data on specific causes of death or serial glucose changes during follow-up. They emphasize that future prospective studies are needed to validate these findings and clarify specific mortality causes.