Characteristics of slow progression to diabetes in multiple islet autoantibody-positive individuals from five longitudinal cohorts: the SNAIL study

Immune-mediated destruction of pancreatic beta cells progresses at different speeds in different individuals. Prospective birth cohort studies show that autoantibodies to insulin (IAA), GAD (GADA), islet antigen-2 (IA-2A), and zinc transporter 8 (ZnT8A) can be detected in children at risk of type 1 diabetes from 6 months of age with a peak in seroconversion between 2 and 3 years [1]. Over 90% of individuals diagnosed in childhood have at least one of these autoantibodies [2]. Presence of multiple islet autoantibodies (mAabs) is normally associated with a disease risk of 70% within 10 years [2‐5], but many individuals present with clinical symptoms later in adult life [1].

Conceptually, disease pathogenesis can be divided into three stages: development of autoimmunity as indicated by the presence of mAabs, defined recently as Stage 1; progression from autoimmunity to glucose intolerance (Stage 2); and finally to overt disease (Stage 3) [6]. These stages appear to be controlled by different genetic mechanisms, with different HLA alleles affecting seroconversion and/or progression to disease [7, 8]. Positivity for and/or higher levels of IAA, IA-2A, IA-2βA and ZnT8A have also been associated with more rapid development of hyperglycaemia [9‐11].

The Slow or Non progressive Autoimmunity to the Islets of Langerhans (SNAIL) study focuses on slow progression to diabetes by studying a large international collection of individuals who develop islet autoimmunity (defined by presence of mAabs) but do not develop disease for at least 10 years. The characteristics of humoral autoimmunity in the earliest mAab-positive sample and the HLA class II profile in this unique ‘slow progressor’ cohort are described here. SNAIL participants derive from BABYDIAB [12], the Diabetes Autoimmunity Study in the Young (DAISY) [13], All Babies in Southeast Sweden (ABIS) [14], the Bart’s Oxford (BOX) Family Study [15] and the Pittsburgh Family Study [5]. These studies all investigated the natural history of diabetes and the contribution of islet autoantibodies to disease risk. Three studies (BABYDIAB, DAISY and ABIS) followed children from birth (with blood samples taken at 9 months or 1 year), while two studies enrolled first degree relatives of individuals with diabetes throughout life. Slow progressors may provide new insights that inform disease prevention strategies in those at risk of type 1 diabetes and help to identify biomarkers associated with slow progression.

Overall, 132 participants were identified who remained diabetes-free for more than 10 years after mAabs were first detected (Table 1). After this 10-year period, participants remained under follow-up (median 4 years, IQR 2–9 years). During follow-up in SNAIL, 42 slow progressors were diagnosed with diabetes, but 90 remained diabetes-free based on lack of self-reported disease, absence from a diabetes case registry or lack of metabolic abnormalities.

Here we describe a unique group of 132 slow progressors with mAab positivity who remained diabetes-free for at least 10 years while participating in five international natural history studies. Most remained diabetes-free at last contact but could be expected to develop diabetes in the future. Indeed, 32% developed diabetes during 4 years of SNAIL follow-up. Diabetes diagnosis was self-reported for the BOX and Pittsburgh studies. This analysis brought together data from several well-characterised long running studies which will enable further analysis of these rare individuals. Most study participants were first-degree relatives of people with type 1 diabetes but also included individuals selected for HLA risk and from the general population.

All five ‘parent’ studies are longitudinal in nature; serum samples were collected and assayed over decades. Slow progressors were initially identified from historic antibody assay screening results. To allow comparison of data, where possible, the initial antibody-positive sample and/or subsequent samples were retested using standardised assays to confirm autoantibody status and these results are reported here.

The definition of slow progressor was the same for all cohorts included in this analysis, although the parent studies had different structures. The BOX and Pittsburgh family studies recruited relatives of individuals with diabetes at different ages and the age of seroconversion is often unknown. In contrast, BABYDIAB, ABIS and most DAISY participants have been studied from birth and provide valuable evidence that even individuals who develop antibodies early in life can remain diabetes-free for many years. Diabetes was defined by clinical disease, but metabolic abnormalities may appear before overt disease. All participants from BABYDIAB were normoglycaemic, while 1 of 10 DAISY participants tested, and 3 of 11 BOX participants tested recently, have elevated HbA_1c. In BABYDIAB the slow progressors described here represent 25% of the children who became mAab-positive in the study [24]. Together these cohorts cover all stages in the natural history of islet autoimmunity; however, individuals with a family history of disease are better represented. The only cohorts included that are representative of the general population are the ABIS study where individuals with no family history were recruited and DAISY where some were recruited from the general population after genetic screening.

All SNAIL participants had mAabs, but the proportion of individuals with each autoantibody varied between study cohorts. The first antibodies to be detected in about two-thirds of young children in BABYDIAB were IAA and their loss was associated with delayed progression [25]. In the SNAIL cohorts, IAA were more common in the first mAab-positive samples from BABYDIAB and ABIS children, but half of BOX and Pittsburgh family study participants were also IAA-positive in their first sample, despite most being tested first as adults. Autoantibodies to GAD are the first islet autoantibody detected in about a third of children who develop diabetes but are also prevalent in adult onset disease [21, 26]. In SNAIL participants, GADA were common in all cohorts but more frequent in older individuals. Antibodies that recognise IA-2 and ZnT8 develop later in disease pathogenesis and are associated with progression [11]; however, ZnT8A were the second most frequent antibodies found in slow progressors. In contrast, IA-2A were less common and BABYDIAB participants had a particularly low prevalence of IA-2A (14%). During follow-up, however, 18 of 22 (82%) BABYDIAB SNAIL participants seroconverted to IA-2A positivity, indicating that antigen spreading continued despite slower progression (data not shown).

The slow progressors in SNAIL have a lower prevalence of DQ2/DQ8 compared with well-characterised rapid progressors diagnosed under the age of five from BOX but a similar genetic risk profile to individuals diagnosed in adolescence. In contrast, BABYDIAB showed that children who develop islet autoantibodies early in life have similar high HLA class II risk, independent of whether they progress to diabetes [24]. One reason for this discrepancy may be that the average age of detection of islet autoantibodies in SNAIL participants was during adolescence. Later development of autoantibodies is associated with lower HLA class II risk as these genes are known to influence antibody development, whereas HLA class I genes have a more dominant role in progression from autoimmunity to disease [7, 8]. The frequency of DQ2 and DQ8 did not differ between cohorts despite differences in participant age between the studies, but this may have been due to a lack of statistical power for comparisons between cohorts. The different structures, ages or ethnicities of the SNAIL cohorts should be considered in interpretation of our analyses. HLA class II risk in children diagnosed under 5 years old is likely to be similar in Western populations, i.e. 36–42% have the highest HLA risk in the type 1 diabetes genetics consortium, DAISY and BOX studies [27]. In addition, over 95% of participants in BABYDIAB, ABIS, BOX and Pittsburgh were of white European origin. For older participants, the age of multiple antibody seroconversion is not known, limiting the interpretation of HLA class II data. Those who seroconverted at a younger age however appear to have similar HLA class II regardless of future progression rate.

On the basis of their mAab positivity, slow progressors are at high risk of developing type 1 diabetes. Despite having a lower genetic risk of disease than rapid progressors, almost 30% of slow progressors carry the highest HLA class II risk haplotype, highlighting the importance of other factors in influencing progression to clinical disease. Continued study of SNAIL participants will focus on characteristics of the immune response, including antibody characteristics, and investigate whether these individuals develop the disease slowly or not at all because of immune regulation.