Background Non-invasive methods of measuring cardiac output are highly desirable in pulmonary arterial hypertension (PAH). We therefore sought to validate impedance cardiography (ICG) against thermodilution (TD) and cardiac magnetic resonance (CMR) in the measurement of cardiac output in patients under investigation for PAH. Methods A prospective, cross-sectional study was performed to compare single-point measurements of cardiac output obtained by impedance cardiography (COICG) technology (PhysioFlow®) with (i) contemporaneous TD measurements (COTD) at rest and steady-state exercise during right heart catheterization and (ii) CMR measurements (COCMR) at rest obtained within 72 h. Results Paired COICG and COTD measurements were obtained in 25 subjects at rest and 16 subjects at exercise. COCMR measurements were obtained in 16 subjects at rest. There was unsatisfactory correlation and agreement between COICG and COTD at rest (r = 0·42, P = 0·035; bias: 1·21 l min−1, 95% CI: −2·33 to 4·75 l min−1) and exercise (r = .65, P = .007; bias: 1·41 l min−1; 95% CI: −3·99 to 6·81 l min−1) and in the change in COICG and COTD from rest to exercise (r = 0·53, P = 0·033; bias: 0·76 l min−1, 95% CI: −3·74 to 5·26 l min−1). There was also a lack of correlation and unsatisfactory agreement between resting COICG and COCMR (r = 0·38, P = 0·1; bias: 1·40 l min−1, 95% CI: −2·48 to 5·28 l min−1). In contrast, there was close correlation and agreement between resting COTD and COCMR (r = 0·87, P<0·001; bias: −0·16 l min−1, 95% CI: −1·97 to 1·65). Conclusions In a representative population of patients under investigation for PAH, ICG showed insufficient qualitative and quantitative value in the measurement of resting and exercise cardiac output when compared with TD and CMR.