We investigated the association of undiagnosed diabetes, previously known diabetes and prediabetes (WHO 1999 classification) with all-cause and cause-specific mortality in an older German population. Previous study results for mortality in patients with very low levels of HbA1c, fasting plasma glucose (FPG), and 2-h plasma glucose (2hPG) are still inconclusive. Thus we have extended the analyses to continuous measures of glycemia. A total of 1,466 subjects aged 55-74 years from the population-based KORA survey S4 (conducted from 1999 to 2001) were included in our observational mortality study (152 subjects with previously known diabetes, and 1,314 further subjects who underwent oral glucose tolerance tests). Mortality was followed up for a maximum of 10.0 years (median follow-up 8.8 years). A total of 180 (12.3%) of the 1,466 subjects have died during the follow-up period. The age- and sex-adjusted hazard ratios for all-cause mortality were 2.6 (95%CI, 1.7-3.8) for known diabetes, 2.8 (95%CI, 1.7-4.4) for undiagnosed diabetes, and 1.1 (95%CI, 0.8-1.7) for prediabetes [reference: normal glucose tolerance (NGT)]. After multivariable adjustment, undiagnosed diabetes was associated with 3.0-fold increased cancer mortality, 1.1-fold increased cardiovascular mortality, and 4.7-fold increased non-cancer, non-cardiovascular mortality compared with NGT. For HbA1c, FPG, and 2hPG, J-shaped associations with all-cause mortality were observed. Undiagnosed diabetes is associated with increased all-cause, cancer, and non-cancer non-cardiovascular mortality, but not with cardiovascular mortality in this older population. All-cause mortality in undiagnosed diabetes is similar to that in previously known diabetes but much higher than mortality in prediabetes and NGT.