The aim of the study was 1) to establish the prevalence of GAD antibodies (GADab) in a population-based study of type 2 diabetes in western Finland, 2) to genetically and phenotypically characterize this subgroup, and 3) to provide a definition for latent autoimmune diabetes in adults (LADA). The prevalence of GADab was 9.3% among 1,122 type 2 diabetic patients, 3.6% among 558 impaired glucose tolerance (IGT) subjects, and 4.4% among 383 nondiabetic control subjects. Islet antigen 2 antibodies (IA2ab) or islet cell antibodies were detected in only 0.5% of the GADab- patients. The GADab+ patients had lower fasting C-peptide concentrations (median [interquartile range]: 0.46 [0.45] vs. 0.62 [0.44] nmol/l, P = 0.0002) and lower insulin response to oral glucose compared with GADab- patients. With respect to features of the metabolic syndrome, the GADab+ patients had lower systolic (140 [29.1] vs. 148 [26.0] mmHg, P = 0.009) and diastolic (79.2 [17.6] vs. 81.0 [13.1] mmHg, P = 0.030) blood pressure values, as well as lower triglyceride concentrations (1.40 [1.18] vs. 1.75 [1.25] mmol/l, P = 0.003). GADab+ men had a lower waist-to-hip ratio compared with GADab- patients. Compared with GADab- patients and control subjects, the GADab+ patients had an increased frequency HLA-DQB1*0201/0302 (13 vs. 4%; P = 0.002) and other genotypes containing the *0302 allele (22 vs. 12%; P = 0.010). However, the frequency of these high-risk genotypes was significantly lower in GADab+ type 2 patients than in type 1 diabetes of young or adult onset (0201/0302 or 0302/X: 36 vs. 66 vs. 64%, P < 0.001). The GADab+ type 2 group did not differ from control subjects with respect to genotypes containing the protective DQB1-alleles *0602 or *0603, nor with respect to the type 1 high-risk genotype in the IDDM1 (Hph1 +/+). We conclude that GADab+ patients differ from both GADab- type 2 diabetic patients and type 1 diabetic patients with respect to beta-cell function, features of the metabolic syndrome, and type 1 diabetes susceptibility genes. Further, we propose that LADA be defined as GADab positivity (>5 relative units) in patients older than 35 years at onset of type 2 diabetes.