External validation of the GEIST score for short- and long-term risk stratification in Takotsubo syndrome: insights from the RETAKO registry

Background/Introduction Takotsubo syndrome (TTS) is associated with significant morbidity and mortality. The GEIST score is one of the few TTS-specific prognostic tools described and has been proposed to predict in-hospital complications and, secondarily, long-term outcomes. It was developed based on 1007 patients from the GEIST registry and externally validated with 946 patients from the RETAKO registry. Further external validation in larger, contemporary cohorts is required. Purpose The aim was to evaluate the prognostic performance of the GEIST score for predicting short and long-term outcomes in the expanded Spanish RETAKO national registry, including a substantially larger cohort with extended follow-up. Methods We conducted a retrospective analysis of 2,174 patients with TTS included in the RETAKO registry (until May 2025). GEIST was calculated at admission. Variables considered incorporated male sex, history of neurological disease, right ventricular involvement, and left ventricular ejection fraction. Its prognostic performance was evaluated for each predefined endpoint based on its corresponding components and time frame. The primary endpoint was a composite of in-hospital complications (death, acute pulmonary oedema, invasive mechanical ventilation, cardiogenic shock), and the secondary endpoint was 5-year all-cause mortality. Discrimination was assessed using ROC curves (AUC), and optimal cut-offs were determined by Youden’s index. Associations between score values above the cut-off points and the outcome were expressed as OR. Finally, time-to-event outcomes were analysed using Cox proportional hazards models and reported as HR; survival was depicted with Kaplan-Meier curves and compared by log-rank test. Results Baseline clinical characteristics are summarized in Table 1, stratified by the selected GEIST cut-off, with the variables considered in the score underlined. Composite in-hospital complications occurred in 493 patients (22.7%), and 277 patients (12.7%) died at 5 years. GEIST demonstrated moderate discrimination for the composite in-hospital endpoint (AUC 0.704), with an optimal cut-off of −33.5 (sensitivity 61%, specificity 72%). At this threshold, the association with in-hospital complications was not statistically robust in adjusted analysis (OR 0.92, 95% CI 0.47–1.81; p=0.804). For long-term prognosis, GEIST performed poorly (AUC 0.566); optimal cut-off was −42.0 (sensitivity 70%, specificity 69%) and no statistically significant association was found with 5-year mortality (OR 1.09, 95% CI 0.80–1.50; p=0.609; Cox HR 1.30, 95% CI 0.99–1.71; p=0.064). Conclusions In this large, national, contemporary TTS cohort, the GEIST score achieved moderate discrimination for in-hospital complications but did not reliably predict long-term mortality. These findings support its use for short-term risk stratification during the acute phase, while highlighting its apparently limited utility for long-term prognostication.