Thirty-eight studies, with 47 study groups, were included in the review (>33,976 patients), including 15 RCTs, five prospective cohorts and 27 retrospective studies. The study sample size ranged from 25 to 6,454 patients. Thirty-six studies reported time in therapeutic range; 18 studies reported international normalised ratio (INR).
The time in therapeutic range ranged from 29 to 75% and was statistically significantly correlated with percentage of INRs in range (r=0.99, p<0.001). Compared with retrospective studies, RCTs had better INR control (65% versus 56%). The time in therapeutic range was statistically significantly negatively correlated with thromboembolic rates (r=-0.59) and major haemorrhage (r=-0.59). A 7% improvement in the time in therapeutic range would lead to reduction of one haemorrhage per 100 patient years (retrospective studies only); a 12% improvement of time in therapeutic range would lead to reduction of one thromboembolic event per 100 patient years (retrospective studies only).
There was no evidence of publication bias.