A few studies have reported a correlation between magnesium and co-morbidity and mortality in end-stage renal disease. We investigated the prognostic value of serum magnesium concentration for mortality in 515 patients on maintenance hemodialysis (60 +/- 12 years, 306 males and 209 females; 24% diabetics). The patients underwent follow-up for 51 +/- 17 (mean +/- SD) months, and the relationship between the baseline magnesium concentration (mean of four months) and outcomes was analyzed statistically. During the follow-up period, there were 103 all-cause deaths, including 63 non-cardiovascular deaths. Kaplan-Meier analysis revealed that mortality was significantly higher in the lower magnesium group (< 2.77 mg/dL, i.e. < 1.14 mmol/L, n = 261), compared to that in the higher magnesium group (> or = 2.77 mg/dL, n = 254) (p < 0.001). Multivariate Cox proportional hazard analysis demonstrated that serum magnesium was a significant predictor for mortality (HR [per 1 mg/dL increase], 0.485 [95% CI, 0.241-0.975], p = 0.0424), particularly for non-cardiovascular mortality (HR 0.318 [95% CI, 0.132 to 0.769], p = 0.0110), after adjustment for other confounders, such as age, gender, hemodialysis duration, and the presence of diabetes. In conclusion, it is demonstrated that lower serum magnesium level is a significant predictor for mortality in hemodialysis patients, particularly for non-cardiovascular mortality, although the mechanisms remain to be explored in future studies. Factors affecting serum magnesium concentrations should be investigated in terms of better survival, including dietary magnesium intake. Further extensive studies may be also needed for possible reconsideration of the current dialysate magnesium concentration (1.0 mEq/L, i.e. 0.50 mmol/L used in most countries), one of the strong contributors to the serum magnesium concentrations of dialysis patients.