Page 38 - Biomarkers for risk stratification and guidance in heart failure

Multimarker risk score in emergency department dyspnea
unselected population with dyspnea at the ED. One recent study combined BNP,
cTnI and hs-CRP in acute heart failure patients and found a gradual increased risk
of 31-day mortality with an increasing number of elevated biomarkers (4.3%, 10%,
20.9% and 53.5%, for 0, 1, 2 or 3 elevated biomarkers respectively)13. Other reports
on combined biomarkers in acute HF or ED dyspnea have focused on long-term 2 rather than short-term risk prediction8, 11, 35 or have focused mainly on the additive
value of one marker on top of NT-proBNP9, 12. Moreover, early differentiation of risk categories is lacking in most previous studies, whereas our MARKED-risk score provides accurate short-term as well as long-term stratification in distinct risk categories, which is crucial for clinical decision-making. The apparent clinical prognostic uncertainty in our cohort - reflected by the relative high admission rate in the low-risk category and relative high discharge rate in the high-risk category with similar mortality rates between admitted and discharged patients – supports the need for accurate risk stratification. Thus, although our study was not designed to assess any therapeutic consequence, objective stratification using the MARKED-risk score into very low, intermediate and high risk may help the treating physician at the ED to decide on urge of intervention, admission, and timing of re-evaluation. However, a clear-cut treatment advice cannot be given based on our score.
We have chosen to use dichotomized values rather than continuous variables to make the score useful for clinical practice. We acknowledge that using cut- points can result in loss of predictive power, but nonetheless our models and score had excellent prognostic accuracy, discrimination and calibration. Other risk scores that are incorporated in practice guidelines such as CHA2DS2–VASc, HAS-BLED, and TIMI3, 48-50 consist of single cut-point variables as well. Furthermore, the predictive accuracy of our score is at least comparable - if not higher - to those of currently used risk scores in other fields (AUC MARKED 0.85, CHA2DS2–VASc, 0.61, HAS-BLED 0.72). Finally, we assessed the value of our multi marker approach on top of the PRIDE-mortality score and found that adding hs-CRP, hs-cTnT and Cys-C to the PRIDE score significantly improved the prognostic accuracy in terms of AUC for both 90-day and 1-year mortality. It should be noted that the PRIDE mortality score was actually developed for 1-year outcome whereas we focus on short-term outcome. Nevertheless, the AUC for 1-year mortality of the PRIDE mortality score in our cohort (0.72, 95% CI 0.68–0.77) was very comparable to the AUC in PRIDE’s validation cohort (0.73, 95% CI 0.64-0.82), indicating that our cohort is representative for cohorts with dyspnea at the ED. Moreover, severity
37