AIMS/HYPOTHESIS: Established autoimmune markers of type 1 diabetes, including islet cell autoantibodies (ICA) and autoantibodies to glutamic acid decarboxylase (GADA) have been used to screen people presenting with type 2 diabetes for latent autoimmune diabetes in adults. We have examined the prevalence of autoantibodies to protein tyrosine phosphatase isoforms IA-2 (IA-2A) and IA-2beta/phogrin (IA-2betaA) in a cohort of adult UKPDS patients thought to have type 2 diabetes, and investigated the possible role of these autoantibodies in predicting requirement for insulin therapy. METHODS: IA-2A and IA-2betaA were measured by a validated radioimmunoassay with human recombinant autoantigens in 4,169 white Caucasian patients aged 25-65 years and newly diagnosed with type 2 diabetes. The clinical requirement for insulin therapy within 6 years was examined in 2,556 patients not randomised to insulin. RESULTS: IA-2A and IA-2betaA were present in 2.2 and 1.4%, respectively, of these patients. IA-2A were more prevalent in younger patients (p for trend <0.00001), more often associated with the HLA-DR4 allele (26.3 vs 8.0%, p<0.0001), and their presence increased the likelihood of insulin therapy requirement within 6 years from diagnosis [relative risk (95%CI) 12.2 (9.8-15.3)]. The presence of IA-2A together with GADA increased the relative risk of requiring insulin therapy from 5.4 (4.1-7.1) for GADA alone to 8.3 (3.7-18.8) and the corresponding positive predictive value from 33 to 50%. CONCLUSIONS/INTERPRETATION: In type 2 diabetes, the presence of IA-2A is infrequent, associated with the HLA-DR4 haplotype, and highly predictive of future need for insulin therapy. The measurement of IA-2betaA does not provide additional information.