Prognostic value of SPECT myocardial perfusion entropy in high-risk type 2 diabetic patients : The EVADIAB study

Background Single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI) for the assessment of myocardial ischemia provides valuable prognostic value. However, the risk stratification of patients with type 2 diabetes mellitus (T2D) remains suboptimal. Recent studies have highlighted the role of coronary microvascular disease in the occurrence of coronary symptoms in this population. Objectives We hypothesized that enhanced signal analysis from SPECT myocardial perfusion images may allow insights into the status of the coronary microvasculature. More specifically, we hypothesized that myocardial perfusion entropy (MPE) as a surrogate marker of microvascular disease quantified from SPECT myocardial perfusion images may provide an incremental prognostic value in T2D patients independently from the routinely performed assessment of myocardial ischemia. Methods T2D patients with very high cardiovascular risk were studied (n = 166, 65 ± 12 years). Ischemia was assessed by SPECT myocardial perfusion imaging (MPI). In addition, SPECT MPI was used for the quantification of rest and stress MPE using an original algorithm providing a global value of MPE from reconstructed rest and stress SPECT images. The primary end point was major adverse cardiac events (MACEs) defined as cardiac death, Q-wave myocardial infarction (MI) and myocardial revascularization > 3 months after SPECT. Results Forty-six patients underwent MACEs over a median follow-up of 4.6 years. Significant differences in stress MPE were observed between patients with and without MACEs (4.19 ± 0.46 vs. 3.93 ± 0.39; P ≤ 0.01). By Kaplan-Meier analysis, the risk of MACEs was significantly higher in patients with higher stress MPE (log-rank P ≤ 0.01). Stress MPE was significantly associated with the risk of MACEs (hazard ratio: 2.80, P ≤ 0.01) after adjustment for clinical and imaging risk factors as identified from preliminary, univariate analysis and including age, hypertension and ischemia. The risk of overfitting was controlled by internal validation procedures such as LASSO, stepwise and random forests selections on time-to-MACEs Cox models. A competing risk event (“other deaths”) slightly overestimated Cox estimates. However, the impact of informative censorship was assessed by Fine-Gray subdistribution hazard estimates and proved to be marginal. The incremental prognostic value of MPE over clinical risk factors was quantified using nested models showing improved AIC, reclassification (global continuous Net Reclassification Improvement [NRI]: 70.1, global Integrated Discrimination Improvement [IDI]: 6.6%), discrimination (change in c-statistic: 0.69 vs. 0.74), and time-dependent AUC (0.71 vs. 0.77 at 4 years of follow-up). Conclusions Stress MPE provided independent and incremental prognostic information for the prediction of MACEs in diabetic patients.