Background During manual chest compressions for cardiac arrest the waveforms of chest compressions are generally sinusoidal, whereas mechanical chest compression devices can have different waveforms, including trapezoidal. We studied the haemodynamic differences of such waveforms in a porcine model of cardiac arrest. Methods Eight domestic pigs (weight 31 ± 3 kg) were anaesthetised and instrumented to continuously monitor aortic (AP) and right atrial pressure (RAP), carotid (CF) and cerebral cortical microcirculation blood flow (CCF). Coronary perfusion pressure (CPP) was calculated as the maximal difference between AP and RAP during diastole or decompression phase. After 4 min of electrically induced ventricular fibrillation, mechanical chest compressions were performed with four different waveforms in a factorial design, and in randomized sequence for 3 min each. Resulting differences are presented as mean with 95% confidence intervals. Results Mean AP and RAP were higher with trapezoid than sinusoid chest compressions, difference 5.7 (0.7, 11) and 6.3 (2.1, 11) mmHg, respectively. Flow measured as CF and CCF was also improved with trapezoidal waveform, difference 14 (2.8, 26) ml/min and 11 (5.6, 17)% of baseline, respectively, with a parallel, non-significant ( P = 0.08) trend for CPP. Active vs. passive decompression to zero level improved CF, but without even a trend for CPP. Conclusion Trapezoid chest compressions and active decompression to zero level improved blood flow to the brain. The compression waveform is an additional factor to consider when comparing mechanical and manual chest compressions and when comparing different compression devices.