Diabetes Care Volume 39, September 2016 1535 b Kendra Vehik,1 Kristian F. Lynch,1 Reversion of -Cell Autoimmunity Desmond A. Schatz,2 Beena Akolkar,3 William Hagopian,4 Marian Rewers,5 Changes Risk of Type 1 Diabetes: Jin-Xiong She,6 Olli Simell,7 Jorma Toppari,7 Anette-G. Ziegler,8 TEDDY Study Ake˚ Lernmark,9 Ezio Bonifacio,10,11 1 Diabetes Care 2016;39:1535–1542 | DOI: 10.2337/dc16-0181 Jeffrey P. Krischer, and the TEDDY Study Group* EPIDEMIOLOGY/HEALTH SERVICES RESEARCH OBJECTIVE b-Cell autoantibodies are a feature of the preclinical phase of type 1 diabetes. Here, we asked how frequently they revert in a cohort of children at risk for type 1 1 diabetes and whether reversion has any effect on type 1 diabetes risk. Health Informatics Institute, Department of Pe- diatrics, Morsani College of Medicine, University RESEARCH DESIGN AND METHODS of South Florida, Tampa, FL 2Department of Pediatrics, College of Medicine, Children were up to 10 years of age and screened more than once for insulin University of Florida, Gainesville, FL autoantibody, GAD antibody, and insulinoma antigen-2 antibodies. Persistent 3National Institute of Diabetes and Digestive autoantibody was defined as an autoantibody present on two or more consecu- and Kidney Diseases, Bethesda, MD 4 fi fi fi Paci c Northwest Diabetes Research Institute, tive visits and con rmed in two reference laboratories. Reversion was de ned as Seattle, WA two or more consecutive negative visits after persistence. Time-dependent Cox 5Barbara Davis Center for Childhood Diabetes, regression was used to examine how reversion modified the risk of development University of Colorado, Aurora, CO 6 of multiple autoantibodies and type 1 diabetes. Center for Biotechnology and Genomic Medi- cine, Medical College of Georgia, Georgia Re- RESULTS gents University, Augusta, GA 7Department of Pediatrics, Turku University Reversion was relatively frequent for autoantibodies to GAD65 (19%) and insulin Hospital, Turku, Finland (29%), but was largely restricted to children who had single autoantibodies (24%) 8Institute of Diabetes Research, Helmholtz Zen- and rare in children who had developed multiple autoantibodies (<1%). Most trum Munchen,¨ and Klinikum rechts der Isar, Technische Universitat¨ Munchen,¨ and Forscher- (85%) reversion of single autoantibodies occurred within 2 years of seroconver- gruppe Diabetes e.V. Neuherberg, Neuherberg, sion. Reversion was associated with HLA genotype, age, and decreasing titer. Germany Children who reverted from single autoantibodies to autoantibody negative 9Department of Clinical Sciences, Lund Univer- ˚ had, from birth, a risk for type 1 diabetes of 0.14 per 100 person-years; children sity/CRC, Skane University, Malmo,¨ Sweden 10Center for Regenerative Therapies, University who never developed autoantibodies, 0.06 per 100 person-years; and, children of Technology, Dresden, Germany who remained single-autoantibody positive, 1.8 per 100 person-years. 11Paul Langerhans Institute Dresden, German Center for Diabetes Research (DZD), Dresden, CONCLUSIONS Germany Type 1 diabetes risk remained high in children who had developed multiple b-cell Corresponding author: Kendra Vehik, kendra. autoantibodies even when individual autoantibodies reverted. We suggest that [email protected]. monitoring children with single autoantibodies for at least 1 year after serocon- Received 27 January 2016 and accepted 20 May 2016. version is beneficial for stratification of type 1 diabetes risk. This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/ b-Cell autoantibodies are significant predictors of type 1 diabetes risk (1,2). The suppl/doi:10.2337/dc16-0181/-/DC1. presence of two or more autoantibodies (insulin autoantibody [IAA], GAD antibody *A complete list of the TEDDY Study Group can [GADA], insulinoma antigen-2 [IA-2A], and zinc transporter type 8 autoantibodies) be found in the Supplementary Data online. and the associated titer have been shown to confer the highest risk of type 1 di- © 2016 by the American Diabetes Association. abetes (3,4). b-Cell autoantibodies develop before type 1 diabetes, are still detect- Readers may use this article as long as the work is properly cited, the use is educational and not able on onset of clinical diabetes, and usually persist over time in the progression to for profit, and the work is not altered. More in- type 1 diabetes. However, b-cell autoantibody titers can fluctuate, and some auto- formation is available at http://diabetesjournals antibody-positive individuals can revert to autoantibody negative (5,6). .org/site/license. 1536 Autoantibody Reversion in TEDDY Diabetes Care Volume 39, September 2016 Persistent b-cell autoantibodies (i.e., history of type 1 diabetes, as previously seroconversion to persistent b-cell au- positive at consecutive visits) are asso- described (8). toimmunity. The specific autoantibody ciated with high-risk type 1 diabetes HLA Children enrolled are monitored pro- result on the initial date was defined genotypes, and transient autoantibody spectivelyfrom3monthsto15years as the baseline measure. Type 1 diabe- expression with lower genetic risk (4,7). with study visits every 3 months until tes was defined according to American Previous findings in the Finnish Type 1 4 years and every 3 or 6 months there- Diabetes Association criteria for diagno- Diabetes Prediction and Prevention after, depending on autoantibody posi- sis (12). HLA-eligible subjects were ,10 (DIPP) and Diabetes Autoimmunity tivity. All children who are persistently years of age and screened a minimum of Study in the Young (DAISY) studies positive for any autoantibody are mon- four times during follow-up for IAA, showed that ;50% of autoantibody- itored every 3 months until the age of GADA, and IA-2A. Children who were in- positive subjects at a single visit revert 15 years or onset of type 1 diabetes. If eligible for this analysis included those to negative within 2 years (5,7). Al- remission of all autoantibodies occurs at who did not have an autoantibody test though little is known about the effects any time during follow-up for a period of result (n = 55) and were HLA ineligible of the variable presence of b-cell auto- four consecutive visits or 1 year, a fol- (n = 118). Children who developed antibodies on the risk of type 1 diabetes, low-up interval of 6 months becomes type 1 diabetes within 3 months of the transience may be true remission of au- effective. Detailed study design and last autoantibody negative (n =6)orfirst toimmunity, humoral markers only, or methods have been previously pub- positive test result (n = 19) were defined just assay variability. Assessing autoanti- lished (9,10). The protocol was ap- as not developing persistent autoanti- body expression may clarify the natural proved by institutional review boards bodies before type 1 diabetes. Other progression of the disease and assist in at participating centers, and all partici- children who developed type 1 diabetes the identification of factors associated pants provided written informed con- more than 3 months after the last auto- with different progression rates. Such sent before participation in the genetic antibody test (n = 14, min–max time af- knowledge will assist in risk profiling as screening and enrollment. ter last test 0.8–8.3 years) were censored well as in reducing the cost burden of at the time of the last test because it was repeat testing of autoimmunity. b-Cell Autoantibodies not possible to determine persistent au- The aim of this study was to deter- b-Cell autoantibodies IAA, GADA, or toantibodies before diagnosis. Autoanti- mine whether the persistence of b-cell IA-2A were measured in two laborato- body-positive children who were lost to autoantibodies over time could stratify ries by radiobinding assays (9,10). In the follow-up immediately after the persis- the risk for type 1 diabetes and further U.S., all sera were assayed at the Bar- tent sample were censored at the time clarify the natural progression of the dis- bara Davis Center for Childhood Dia- of the initial seroconversion because easeanditsassociationwithdifferent betes at the University of Colorado they were never observed at risk for re- progression rates in The Environmental Denver; in Europe, all sera were assayed version (n = 14). Determinants of Diabetes in the Young at the University of Bristol, U.K. Both Autoantibody titers were converted (TEDDY) study of genetically high-risk laboratories reported high sensitivity, to z-scores due to different autoanti- children. specificity, and concordance (11). All body cutoff values in the two reference positive b-cell autoantibodies and 5% laboratories (Bristol and Denver) and RESEARCH DESIGN AND METHODS of negative samples were retested in change in type of assay over time from Study Population the other reference laboratory and the original TEDDY standard assay to the TEDDY is a prospective cohort study of deemed confirmed if concordant. Auto- harmonized assay. Autoantibody titers children at genetic high risk for type 1 antibody determination of maternal were converted to SD units centered diabetes, funded by the National Insti- acquired or de novo appearance was on the assay threshold for analysis. The tutes of Health, which seeks to identify based on the presence of both maternal harmonized assay was used for IA-2A environmental causes of type 1 diabetes. and child autoantibodies over the first and GADA; if a measure was missing There are six clinical research centersd 18 months of age. De novo production from the harmonized assay, the TEDDY three in the U.S.: Colorado, Georgia/Florida, was designated if the mother was nega- assay transformed to the harmonized Washington, and three in Europe: Finland, tive for autoantibodies and the child assay was used. The Bristol z-score was Germany, and Sweden. The high-risk geno- was positive. If the mother was positive, used for IAA (not harmonized), and, if types for subjects screened from the gen- then the child was deemed negative for missing, the Denver z-score was used.