Plasma homocysteine concentrations have been associated with the risk of stroke, but its relevance to secondary vascular events and mortality after stroke remains unclear because of inconsistent results from clinical trials. The aim of the present study was to investigate whether plasma homocysteine levels and the MTHFR (methylenetetrahydrofolate reductase) variant C677T contributed to the risk of stroke recurrence and all-cause mortality in a large prospective cohort of stroke patients in a Chinese population. A total of 1823 stroke patients (age, 35-74 years) were recruited during 2000-2001 and prospectively followed-up for a median of 4.5 years. During the follow-up, 347 recurrent strokes and 323 deaths from all-causes were documented. After adjustment for age, gender and other cardiovascular risk factors, a high homocysteine concentration was associated with an increased risk of 1.74-fold for stroke recurrence {RR (relative risk), 1.74 [95% CI (confidence interval), 1.3-2.3]; P < 0.0001} and 1.75-fold for all-cause mortality [RR, 1.75 (95% CI, 1.3-2.4); P<0.0001] when highest and lowest categories were compared. Spline regression analyses revealed a threshold level of homocysteine for stroke recurrence. By dichotomizing homocysteine concentrations, the RRs were 1.31 (95 % CI, 1.10-1.61; P = 0.016) for stroke recurrence and 1.47 (95 % CI, 1.15-1.88; P < 0.0001) for all-cause mortality in patients with homocysteine levels >= 16 mu mol/l relative to those with levels < 16 mu mol/l. The association of elevated plasma homocysteine concentrations with all-cause mortality was mainly due to an increased risk of cardiovascular deaths. No significant association was found between MTHFR C677T and stroke recurrence or mortality. In conclusion, our findings suggest that elevated homocysteine concentrations can predict the risk of stroke recurrence and mortality in patients with stroke.