Abstract Background The therapeutic landscape of chronic myeloid leukaemia (CML) has been transformed by tyrosine kinase inhibitors (TKI). Nilotinib, a potent 2nd generation TKI, showed higher rates of major molecular response than imatinib, however was associated with higher rates of cardiovascular (CV) toxicity. Objectives To describe the CV events associated with nilotinib in a real-world CML population and assess the predictive value of the HFA-ICOS baseline risk score in clinical practice. Methods The HFA-ICOS baseline risk was calculated (low, medium, high /very high) for patients with CML treated with nilotinib between 2006 and 2021. The incidence of all CV events, ischaemic events and Major CV Adverse Events (MACE) were reviewed and compared among the different risk groups. Results A total of two hundred and twenty-nine eligible patients were included in the analysis. The incidence of CV events, ischaemic events, and MACE, while on nilotinib, were 20.9% (95% CI: 15.7% to 26.2%), 12.7% (95% CI: 8.4% to 16.9%), and 12.2% (95% CI: 7.9% to 16.5%) respectively, following a median duration of treatment with nilotinib of 34.4 months for the entire cohort. Patients with a higher HFA-ICOS baseline risk score had higher rates of all CV events (low: 11.2%, medium: 28.2% [HR: 2.8, 95% CI: 1.2 to 6.2], high/very high: 32.4% [HR: 3.5, 95% CI: 1.8 to 6.8]), ischaemic events (low: 5.2%, medium: 17.9% [HR: 3.7, 95% CI: 1.2 to 10.9] , high/very high: 21.6% [HR: 3.7, 95% CI: 1.4 to 9.7]) and MACE (low: 7.8% , medium: 10.3% [HR:1.2, 95% CI: 0.38 to 4.1], high/very high: 20.2% [HR: 2.2, 95% CI: 0.9 to 5.3]). The predictive tool presented an 89% NPV, and an area under the ROC curve: 0.65 for all CV events, an NPV of 95% and AUC 0.68 for ischaemic events and NPV 92% and AUC 0.62 for MACE. Conclusions The HFA-ICOS risk stratification tool has been shown to be an efficient discriminator at low, medium and high/very high risk of developing cardiovascular events, with an overall positive trend towards increasing cardiotoxicity rates with rising risk categories. This study provides evidence to support the use of this predictive tool in nilotinib treated patients, as it also underscores the potential for further improvement in the overall discriminatory ability of the model.