Geert Kleinnibbelink

Chapter 6 154 All-cause mortality vs. combined endpoint Our meta-analysis revealed a lower predictive capacity for combined endpoints versus all- cause mortality. This difference may be explained by the fact that clinical events included in the analysis for the combined endpoint are heterogeneous and, therefore, not all events may directly relate to strain (hence, the lower predictive capacity). Other factors than cardiac strain (e.g. gas transfer in the lungs 39 ) may contribute to the occurrence of these clinical events. In addition, several studies included intensified PH medical therapy as a combined endpoint, whilst this unlikely relates to cardiac strain. Therefore, the diversity in clinical events included in the combined endpoint, but also the weak link between some of these factors and cardiac strain, lowers the discriminating capacity of RVLS to predict a combined endpoint versus all-cause mortality. Predictive capacity vs. a relative reduction in % of RVLS As shown in Figure 3 there is no clear relation between the relative reduction in % of RVLS and the HR for all-cause mortality. This may be explained by the differences across study designs. In contrast to our expectations, the three studies with the lowest relative reduction in RVLS presented the highest HRs in the analysis for all-cause mortality. These three studies all used a dichotomous cut-off value (between -17% and -20%) for RVLS 9, 15, 28 , which was higher than the mean RVLS value for the PH patients in the two remaining studies (i.e. -16.1% and -15%). 10, 23 The latter two studies calculated the HR per SD-unit change in RVLS, which resulted in a lower absolute cut-off (approximately -11.1 and -10%) value and in a higher incidence of mortality in the group above the cut-off value. In contrast to the cut-off values in the latter two studies, additional analysis to identify the ideal cut-off value in 4 out of these 5 studies showed that an absolute cut-off between -12.5% and -19.1% had the highest sensitivity and specificity to detect all-cause mortality in PH patients. 9, 10, 23, 28 This indicates that the calculated HR per SD-unit change underestimates the predictive value of RVLS in the latter two studies. Future direction and clinical implications Outcomes of the present meta-analysis supports the use of RVLS in patients with PH. Although RVLS has independent predictive value, recent strategies for predicting mortality and events in PH patients consists of constructing multi-parameter predictive