Geert Kleinnibbelink

RV Strain in Pulmonary Hypertension 6 145 Data extraction Data was independently extracted by two authors (H.H. and G.K.) using a predetermined data extraction file. Differences in data extraction were resolved by consensus and if necessary a third author was consulted (T.E.). Since all selected studies included strain, but only one study stain and strain rate, we focused on the prognostic value of strain only. Univariate and multivariate HR (95%-CI), the mean RVLS for the study population and the RVLS cutoff value for calculation of the HR were extracted from the individual studies ( Table 2 ). The included studies reported HRs on either a continuous scale (i.e. change in risk per % RVLS) and/or a dichotomous scale (i.e. below/above a cut-off point). In case of a dichotomous scale theHR should increasewithahigher absolutevalue (due to thenegative nature of RVLS), but as some studies investigated the beneficial effect of a RVLS value below a certain cut-off point, we calculated the inverse HR (1/HR – [1/95%-CI]) to ensure homogeneous presentation of the data. Additional information gathered consisted of: sample size, age, sex, World Health Organisation (WHO) class, New York Heart Association (NYHA) class, the follow-upperiod and the clinical endpoint of the individual studies ( Table 3 ). For assessment of study quality, data regarding the echocardiographic assessment was gathered consisting of manufacturer, assessment software, echocardiographic window / image, included segments, methods of optimization and usage of the guidelines. When viable data was missing, an attempt was made to request missing data from the authors by email (n=4 studies). Three out of four studies with missing data provided the requested information and were included in our meta-analysis ( Figure 1 ).