(a) Photograph showing a functioning microfluidic device attached to the femoral vein of a pig via a trilumen catheter. Blood is pulled at a set flow rate (shear gradient=4,375 s−1 mm−1) so that it passes through the device and the inline pressure sensor, forming clots inside the device; pressure data are automatically acquired on a laptop. (b–d) endotoxemic shock model. Lipopolysaccharide endotoxin (LPS) was infused into a pig over 1.5 h (grey shaded region, −1.5–0 h), and then thrombin-antithrombin (TAT) complexes in plasma (b; baseline=11.95 ng ml−1), fibrinogen concentration in plasma (c; baseline=143.61 ng ml−1), and microfluidic clotting time (μCT) calculated using the microchip (d; baseline=20.64 min; *P<0.05 compared with baseline, one-way ANOVA, s.e.m.) were measured every hour thereafter. (e) Pearson correlation coefficient (r) of aPTT (light grey), ACT (dark grey) and μCT (green) compared with TAT and fibrinogen measurements (n=3 pigs, 2 replicates per experiment). (f,g) Heparin therapy model. Fold change (relative to baseline at no heparin), in heparin concentration (black circles) measured by Factor Xa assay (f) and μCT (black circles) measured using the microfluidic haemostasis monitor (g) (baseline=4.25 min; full line, line of linear regression; dotted line, 95% confidence interval of the linear fit). (h) Graph showing the sensitivity of aPTT (light grey), ACT (dark grey) and μCT (green) assays (**P<0.01, one-way ANOVA, s.e.m.; n=3 pigs, 2 replicates per experiment).