LAUSR

Exercise echocardiography (ExE) can predict overall and cardiovascular mortality. ExE provides information from two sources: imaging and exercise electrocardiogram testing. However, the relative contributions of imaging and the ‘exercise part’ of an ExE study for prognostication of the different causes of mortality are less known. We aimed to assess the value of imaging and exercise variables of an ExE for the prediction of the different causes of death. We studied 12,615 consecutive patients with ExE performed between 1995 and 2014. Reasons for ExE included chest pain in 72%. The Bruce protocol was mostly used (93%). Maximal exercise workload in metabolic equivalents (METs) was derived from exercise testing characteristics (slope and speed). A good functional capacity was defined as a maximal workload of 10 METs. Peak treadmill imaging acquisition was performed with the patient exercising, as reported (https://www.youtube.com/watch?v1⁄4gnYRa6PoYVw). Wall motion score index and its change with exercise were calculated. Ischaemia was defined as new/worsening wall motion abnormalities (WMA) with exercise, and fixed WMAs as resting WMAs without changes with exercise. All patients gave informed consent. Follow-up was obtained by hospital databases review, death certificates and telephone interviews. End point was cardiovascular (CV), cancer (CA) or non-cardiovascular non-cancer (NCV-NCA) death. CV death was considered in the case of cardiac death, stroke or complications from arteriosclerosis. Cardiac death was defined as death due to acute myocardial infarction, congestive heart failure, life-threatening arrhythmias or cardiac arrest. Unexpected, otherwise unexplained sudden death was considered cardiac death. NCVNCA death included infectious diseases, neurological, pulmonary, renal failure, liver disease, multiorgan failure, non-cardiac surgery, accident/trauma, suicide, other non-cardiac; and unknown or unobtainable. CA deaths were identified by codes from the International Classification of Diseases 9th/10th revisions. Information on causes of death was provided by the local community registry (Registry of Mortality of Galicia). Continuous variables were reported as mean1 standard deviation and intergroup differences assessed with analysis of variance. Univariable and multivariable associations of variables with the different causes of death were assessed. To compare the subdistribution of hazard ratios for each cause-specific mortality according to different variables, modified Cox regression hazard models were employed by the Fine– Gray method. This test considers as a single cause of death both the association of the different variables with a single cause of death and the contribution of another competing event by actively maintaining subjects in the risk sets. For those variables measuring related aspects, only the one with the higher C-index was included. A cause-specific C-index in the presence of competing risks was considered. The cumulative incidence in competing risk analyses was calculated using the cmprsk package of R. The cause-specific C-index was computed using the C-index function from the risk Regression package of R. A two-sided p value< 0.05 was considered significant.