5 Clinical prediction scores with or without NTproBNP in predicting stroke and transient ischemic attack in at-risk patients without atrial fibrillation – the stop-HF experience

Background Clinical prediction rules, such as CHA2DS2-VASc, HATCH and HAVOC have been used in the prediction of stroke and transient ischemic attack (TIA) in those with atrial fibrillation (AF). Currently the ESC guidelines recommend the use of the CHA2DS2-VASc score to help determine which patients with AF would most benefit from anticoagulation. In recent years there has been a trend to use CHA2DS2-VASc for stroke prediction even in the non-AF population, with similar C-statistic (0.61–0.76) in comparison to AF cohorts. Similarly, natriuretic peptides have gained interest as a marker of cardiovascular risk, including cardioembolic stroke. Purpose We compared CHA2DS2-VASc, HA2TCH and HAVOC for prediction of stroke/TIA in patients without a history of atrial fibrillation or flutter and hypothesised that adding natriuretic peptides would improve clinical risk prediction for stroke or TIA in the STOP-HF population. Methods The study included 1156 patients without known AF enrolled in the STOP HF follow-up study between 2006 and 2017. The end-point measured was Stroke/TIA. Baseline NTproBNP levels(NP), CHA2DS2-VASc, HAVOC and HATCH scores were recorded for all patients. Logistical regression analysis and receiver operating characteristic curve were used to estimate odds ratio and the diagnostic accuracy of the NP and clinical prediction scores for Stroke/TIA, respectively. Results The mean age (SD) was 65.8 (10.1) years, 579 (50.5%) were female, prominent co-morbidities were hypertension and diabetes (table 1). The median NTproBNP was 93 pg/ml (IQR:53- 189). The endpoint of stroke/TIA occurred in 44 (3.8%) of the population. Patients experiencing stroke/TIA over follow up were older, more likely to have baseline vascular and thromboembolic disease and higher baseline NTproBNP levels than those who did not experience stroke/TIA during follow up. The median scores for the CHA2DS2-VASc, HAVOC, HATCH were also higher at baseline higher in the new onset TIA/Stroke group. All the clinical prediction scores as well as NTproBNP predicted Stroke/TIA (table 2). All clinical prediction scores remained significant predictors of Stroke/TIA after multivariate adjustment. NP prediction was no longer significant (table 2). The HATCH score significantly outperformed the other scores with a c statistic of 0.76 (95% CI, 0.69 - 0.78) (figure 1 & 2). The c-statistic for the CHA2DS2-VASc and HAVOC scores were 0.69 (95% CI, 0.61- 0.77) and 0.66 (95% CI, 0.57- 0.74), respectively. Addition of NP did not improve the discriminatory power of any of the scores. Conclusion All the clinical scores predicted Stroke/TIA in patients without known AF. However, the HATCH score significantly outperformed the CHA2DS2-VASc and HAVOC scores. Baseline NP levels, neither predicted Stroke/TIA, nor enhanced the prediction power of the clinical scores. Further evaluation of the HA2TCH score will delineate its potential value in clinical practice.