Clinical determinants of agreement and discordance between stress SPECT and invasive coronary angiography

Background: Stress single-photon emission computed tomography (SPECT) myocardial perfusion imaging offers a non-invasive alternative for invasive coronary angiography (ICA) in diagnosing coronary artery disease (CAD), however often yields inconclusive results. This study aimed to characterize patterns of agreement and discordance between stress SPECT and ICA, and to evaluate potential risk stratification. Methods: We retrospectively analyzed 915 patients with suspected CAD who underwent stress SPECT followed by ICA within three months (enrollment: 2019–2020). Clinical, demographic, and imaging data were extracted from medical records. Variables associated with diagnostic agreement were identified using univariate logistic regression. A predictive model combining SPECT results with clinical variables was developed using backward stepwise logistic regression and evaluated by AUC-ROC. Results: Diagnostic agreement between stress SPECT and ICA was observed in 624 patients (68.2%), while 291 patients (31.8%) demonstrated discordant results. Agreement was associated with use of nitrates (OR 3.18, 95% CI 1.31–7.73), antiplatelet therapy (OR 2.57, 95% CI 1.86–3.56), renal failure (OR 2.34, 95% CI 1.43–3.84), and type II diabetes mellitus (OR 1.78, 95% CI 1.28–2.48), whereas female sex (OR 0.50, 95% CI 0.34–0.73), smoking (OR 0.72, 95% CI 0.50–1.03), and higher body mass index (BMI; OR 0.95 per kg/m², 95% CI 0.92–0.99) were associated with disagreement. The final multivariable model included stress SPECT results, sex, BMI, smoking status, serum creatinine, renal failure, type II diabetes mellitus, and use of antiplatelets and nitrates and demonstrated improved discrimination compared with SPECT alone (AUC 0.72 vs. 0.54, p < 0.001). Conclusion: In this exploratory study, clinical factors were associated with an agreement between stress SPECT and ICA. Incorporating clinical context alongside SPECT findings may help inform risk stratification.