Adult-Onset Type 1 Diabetes: Current Understanding and Challenges

Incidence of Type 1 Diabetes Among Adults Worldwide

Adult-onset type 1 diabetes is more common than childhood-onset type 1 diabetes, as shown from epidemiological data from both high-risk areas such as Northern Europe and low-risk areas such as China (3–8). In southeastern Sweden, the disease incidence among individuals aged 0–19 years is similar to that among individuals 40–100 years of age (37.8 per 100,000 persons per year and 34.0/100,000/year, respectively) (3). Given that the comparable incidence spans only two decades in children, it follows that adult-onset type 1 diabetes is more prevalent. Similarly, analysis of U.S. data from commercially insured individuals demonstrated an overall lower incidence in individuals 20–64 years of age (18.6/100,000/year) than in youth aged 0–19 years (34.3/100,000/year), but the total number of new cases in adults over a 14-year period was 19,174 compared with 13,302 in youth (4). Despite the incidence of childhood-onset type 1 diabetes in China being among the lowest in the world, prevalence data show similar trends across the life span. From 2010–2013, the incidence was 1.93/100,000 among individuals aged 0–14 years and 1.28/100,000 among those 15–29 years of age versus 0.69/100,000 among older adults (5). In aggregate, adults comprised 65.3% of all clinically defined newly diagnosed type 1 diabetes cases in China, which is similar to estimates using genetically stratified data from the population-based UK Biobank using a childhood-onset polygenic genetic risk score (GRS) (6). It is important to note that the proportion would likely be higher if autoimmune cases not requiring insulin initially were classified as type 1 diabetes. For example, in a clinic-based European study, the proportion of adults with diabetes not initially requiring insulin yet with type 1 diabetes–associated autoantibodies was even higher than those started on insulin at diagnosis with a defined type 1 diabetes diagnosis (9). Moreover, in an adult population-based study in China, the fraction (8.6%) with diabetes not requiring insulin yet with type 1 diabetes–associated autoantibodies was similar to that in Europe, implying that there could be over 6 million Chinese with adult-onset type 1 diabetes (10). While there is a wide range in the incidence of type 1 diabetes across different ethnic groups, even using differing methods of case identification (7), these data support the notion that, worldwide, over half of all new-onset type 1 diabetes cases occur in adults.

Natural History Studies of Type 1 Diabetes

Our understanding of the natural history of type 1 diabetes has been informed by a number of longitudinal and cross-sectional studies. At one end of the spectrum are prospective birth cohort studies, such as the BABYDIAB study in Germany and The Environmental Determinants of Diabetes in the Young (TEDDY) study, which includes sites in Germany, Finland, Sweden, and the U.S. While these studies now have the potential to explore the pathogenesis of islet autoimmunity by being extended into adulthood, they have primarily focused on events occurring in childhood (11). Clinical centers in North America, Europe, and Australia collaborate within Type 1 Diabetes TrialNet, a study that identifies autoantibody-positive adults and children in a cross-sectional manner to examine the pathogenesis of type 1 diabetes and to perform clinical trials on those at high risk in order to preserve β-cell function (12). At the other end of the spectrum, the European Prospective Investigation into Cancer and Nutrition (EPIC)-InterAct study is a case-cohort study nested in the U.K. prospective adult population-based EPIC study (13), while the clinical, immunogenetic, and metabolic characteristics of autoimmune adult-onset type 1 diabetes have been extensively studied in large American, European, and Chinese studies, including UK Prospective Diabetes Study (UKPDS), Action LADA, Scandia, Non Insulin Requiring Autoimmune Diabetes (NIRAD), and LADA China (9,14–19). Based on these cross-sectional and prospective studies, considerable data have been generated to define differences within type 1 diabetes according to the age at onset. Here, we highlight key aspects of age-related genetic, immune, and metabolic heterogeneity in type 1 diabetes. Of note, the term latent autoimmune diabetes in adults (LADA) has been used to describe adults with slowly progressive autoimmunity, sometimes exhibiting features overlapping with those of type 2 diabetes (9,14,18). At the outset of the workshop and for the purposes of this Perspective, LADA was not considered a unique entity; rather, we considered the classification of type 1 diabetes to include all individuals with evidence of autoimmunity, regardless of the trajectory of disease development (i.e., rapid or slowly progressive) or other associated demographic and/or clinical features (e.g., obesity).

Age-Related Genetic Heterogeneity

Type 1 diabetes shows heterogeneity across a broad range of clinical, genetic, immune, histological, and metabolic features (20). Childhood-onset type 1 diabetes is most often attributed to susceptibility alleles in human leukocyte antigen (HLA), which contribute ∼50% of the disease heritability. Whereas ethnic differences exist, notably for specific HLA genotypes, several broad principles apply. Compared with childhood-onset disease, adult-onset type 1 diabetes cases show lower type 1 diabetes concordance rates in twins (21), less high-risk HLA heterozygosity (19), lower HLA class I (14), more protective genotypes (14,15), and lower GRS (6,22), which are calculated by summing the odds ratios (OR) for disease-risk alleles.

Diabetes-Associated Immune Changes

Adult-onset type 1 diabetes, like childhood-onset type 1 diabetes, is associated with the presence of serum autoantibodies against β-cell antigens. Serum glutamic acid decarboxylase (GADA) autoantibodies may be useful as a predictor of type 1 diabetes in adults, as adult-onset cases most often present with GADA positivity (9,10,15,17,18,20,22) and possess an HLA-DR3 genotype (9,14,15, 20,21,23). In one prospective study of a general population, the hazard risk of incident diabetes in those with a high type 1 diabetes GRS and GADA positivity was 3.23 compared with all other individuals, suggesting that 1.8% of incident diabetes in adults was attributable to that combination of risk factors (13). In adult-onset type 1 diabetes, multiple diabetes-associated autoantibodies tend to be less prevalent with increasing age at diagnosis (1,8), yet GADA remains the dominant autoantibody irrespective of the need for insulin treatment at diagnosis and irrespective of ethnicity (9,17,18,24,25), even despite a paucity of HLA DR3, as in Japan and China (17,18). In contrast, childhood-onset type 1 diabetes cases often have insulin autoantibodies and an HLA-DR4 genotype, higher identical twin disease concordance, more HLA heterozygosity, and higher GRS (20). Taken together, these data indicate that type 1 diabetes is heterogeneous across the spectrum of diagnoses, suggesting that pathogenesis and optimal therapy are also diverse.

Data from the TrialNet Pathway to Prevention cohort demonstrated lower risk of progression to type 1 diabetes in adults than children, even when both show multiple autoantibodies on a single occasion and are monitored over 10 years (12). One recent analysis found that the 5-year rate of progression to diabetes in multiple autoantibody–positive adults was only ∼15%, with a number of them remaining diabetes-free for decades (26). A combined cohort study, known as the Slow or Nonprogressive Autoimmunity to the Islets of Langerhans (SNAIL) study, is following such “slow progressors” with multiple autoantibodies who have yet to progress to stage 3 type 1 diabetes (i.e., clinical diagnosis) over at least a 10-year period (27). Many of these slow progressors lose disease-associated autoantibodies over time, adding complexity to cross-sectional classification (28). Based on estimates from natural history studies, slow progressors, even if identified when young, cannot account for all autoimmune adult-onset diabetes, indicating that autoantibodies must develop at all ages (11). However, little is known about those who initially develop autoimmunity as adults, mostly due to the lack of longitudinal studies focusing on this population.

People with type 1 diabetes, in contrast to the majority of those with type 2 diabetes, have altered adaptive immunity (i.e., islet autoantibodies and T-cell activation), while innate immune changes, including cytokine changes, are common to both (29). Increased T-cell activation by islet proteins has also been found in a proportion of adults with initially non-insulin-requiring diabetes, even when they lack diabetes autoantibodies (30). However, there is a paucity of immune studies on adult-onset type 1 diabetes and few histologic studies. An analysis of tissues from the Network for Pancreatic Organ Donors with Diabetes (nPOD) showed no relationship between age at diabetes onset and the frequency of islet insulitis (31). The composition of islet insulitis differs in very young children compared with older individuals, with the former having an increased frequency of B cells in islet infiltrates (32). However, relating pancreatic histological changes to changes in peripheral blood remains a challenge.

Adults with new-onset type 1 diabetes are at increased risk of other autoimmune conditions. About 30% of individuals with adult-onset type 1 diabetes have thyroid autoimmunity (27,29). In addition, adults with type 1 diabetes who possess high-titer GADA and/or multiple islet autoantibodies are at increased risk of progression to hypothyroidism (24,33). In a large population-based Chinese study, the prevalence of adult-onset type 1 diabetes was 6% among initially non-insulin-requiring diabetes cases, and 16.3% of them had thyroid autoimmunity (OR 2.4) (10). Of note, those with islet antigen 2 autoantibodies had a high risk of tissue transglutaminase autoantibodies, a marker for celiac disease (OR 19.1) (10). Thus, in the clinical setting, there should be a high index of suspicion for other autoimmune conditions in individuals with adult-onset type diabetes, and associated autoimmunity should be screened where clinically indicated.

Metabolic Characteristics of Adult-Onset Type 1 Diabetes

Age-related differences in type 1 diabetes extend to metabolic parameters. C-peptide at diagnosis is higher in adults than children, driven in part by higher BMI (34). Analysis of U.K., TrialNet, and Chinese cohorts has identified two distinct phases of C-peptide decline in stage 3 disease: an initial exponential fall followed by a period of relative stability. Along with initial differences at the time of clinical diagnosis, the rate of decline over 2–4 years was inversely related to age at onset (10,34–36). Furthermore, the U.S. T1D Exchange Study found that glycemic control was better in adults with type 1 diabetes than in children and adolescents with type 1 diabetes (37). The American Diabetes Association (ADA) targets for glycemia are higher in children, so that in this same cohort, 17% of children, compared with 21% of adults, achieved the ADA hemoglobin A_1c (HbA_1c) goal of <7.5% and <7.0%, respectively (37). Other factors confound this relationship between age at diagnosis and metabolic control. First, individuals with adult-onset type 1 diabetes are more likely to have residual insulin-producing β-cells and persistent measurable C-peptide in disease of long duration, the latter of which has been linked to improved glycemic control (38,39). Second, individuals with adult-onset type 1 diabetes, initially not on insulin therapy, tend to have worse metabolic control than people with type 2 diabetes, even when receiving insulin treatment (9,40). The sole exception is the LADA China study, where worse control was noted only among those with a high GAD titer (18). Metabolic differences between adults and children extend beyond C-peptide. Adults with autoantibody positivity who progressed to type 1 diabetes were less likely than very young children to exhibit elevated proinsulin/C-peptide ratios prior to stage 3 disease onset (41). In addition, in individuals with disease of long duration, those diagnosed at an older age had evidence of improved proinsulin processing and nutrient-induced proinsulin secretory capacity (42).

Diagnosis and Management of Adult-Onset Type 1 Diabetes

Correctly identifying diabetes etiology and type is difficult, and misclassification may occur in up to 40% of adults presenting with type 1 diabetes (1,2). Reasons underlying misclassification are multiple and include 1) lack of awareness that the onset of type 1 diabetes is not limited to children; 2) the overwhelming majority of people developing diabetes as older adults have type 2 diabetes, contributing to a confirmation bias (2); 3) typical clinical criteria, such as BMI and metabolic syndrome, can be poor discriminators, especially as rates of obesity in the overall population increase (9,43); 4) clinical characteristics of adult-onset type 1 diabetes can masquerade as type 2 diabetes, given their slow metabolic progression and risk of metabolic syndrome (which occurs in about 40%), so that the distinction between types of diabetes may be blurred (43–45); and 5) lack of awareness of and accessibility to biomarkers that may serve as tools to distinguish type 1 diabetes and type 2 diabetes.

Tools to distinguish type 1 and type 2 diabetes are under active development. For example, classification models integrating up to five prespecified predictor variables, including clinical features (age of diagnosis and BMI) and clinical biomarkers (autoantibodies and GRS) in a White European population, had high accuracy to identify adults with recently diagnosed diabetes with rapid insulin requirement despite using GRS derived from childhood-onset type 1 diabetes. While GRS have the potential to assist diagnosis of type 1 diabetes in uncertain cases, they are not yet widely available in clinical practice. Moreover, it is important to note that while the model was optimized with the inclusion of all five variables, the addition of GRS had only a modest effect on overall model performance (22).

Classification can be aided by the measurement of autoantibodies and C-peptide. Recommended autoantibodies to assay at the time of diagnosis include those to insulin (insulin autoantibody), glutamate decarboxylase isoform 65 (GAD65A), insulinoma antigen 2, and zinc transporter isoform 8 (Znt8A), with GAD65A being the most prevalent autoantibody among adults. High levels or the presence of more than one antibody increases the likelihood of type 1 diabetes. However, it is important to realize that islet autoantibodies are a continuous marker that can also occur in the population without diabetes. As with many other tests, an abnormal test is usually based on a threshold signal from control populations without diabetes, usually the 97.5th or the 99th centile. Therefore, false-positive results with these assays can occur and can be reduced by using higher-specificity assays or thresholds and targeting testing toward those with clinical features suggestive of type 1 diabetes (46). Finally, since antibody levels can wane over time in established type 1 diabetes, the absence of autoantibodies does not rule out the possibility of a diagnosis of type 1 diabetes.

Measurement of C-peptide, paired with a blood glucose in the same sample, provides an estimate of endogenous insulin production and has the most utility in disease of long duration when levels fall below 300 pmol/L (39,47). However, C-peptide levels are typically higher at presentation and may be difficult to distinguish from levels in type 2 diabetes, which are usually >600 pmol/L. Thus, thresholds of C-peptide that clearly delineate type 1 diabetes from type 2 diabetes at diagnosis cannot be categorically defined, and C-peptide must be interpreted within the context of other clinical and laboratory features. Measurement of a random nonfasting C-peptide is superior to fasting C-peptide in identifying type 1 diabetes (48) and is well correlated with stimulated C-peptide levels measured during a mixed-meal tolerance test, which is considered the gold standard assessment of insulin secretory function in established type 1 diabetes (49). A recent analysis found that concomitant blood glucose ≥144 mg/dL (8 mmol/L) increased the specificity of random C-peptide in predicting a stimulated C-peptide level <600pmol/L, suggesting this is a reasonable threshold of blood glucose to employ for C-peptide interpretation (49).

C-peptide also can be used to guide therapy (50). Individuals with a random C-peptide level ≤300 pmol/L should be managed mainly with insulin. For those with random C-peptide levels >300 pmol/L, insulin could be combined with other diabetes therapies, although evidence about safety and efficacy is limited. It is generally agreed that sulfonylureas should be avoided because of the potential to hasten β-cell failure (50). There is concern for increased risk of diabetic ketoacidosis (DKA) with sodium–glucose cotransporter 1 (SGLT1) and SGLT2 inhibitors when these agents are used in type 1 diabetes, especially in nonobese individuals who may need only low dosages of insulin (51). All other agents could be considered for therapy in those not requiring insulin initially. In individuals with random C-peptide levels exceeding 600 pmol/L, management can be much as recommended for type 2 diabetes, with the caveats outlined above (50). An important consideration is that loss of β-cell function may be rapid in autoimmune diabetes. As such, individuals treated without insulin should be closely monitored.

In the absence of prospectively validated decision support tools that have been tested in multiethnic populations, we suggest, as an approach to aid the practicing physician, assessment of age, autoimmunity, body habitus/BMI, background, control, and comorbidities, using the acronym AABBCC (Table 2). This approach includes the clinical consideration of autoimmunity and other clinical features suggestive of type 1 diabetes, including age at diagnosis, low BMI, an unexplained or rapid worsening of clinical course manifesting as a lack of response or rising HbA_1c with type 2 diabetes medications, and a rapid requirement for insulin therapy, especially within 3 years of diagnosis. It should be emphasized that among these features, age at diagnosis (<40 years), low BMI (<25 kg/m²), and rapid need for insulin therapy are the most discriminatory (43). We recommend measurement of islet antibodies and C-peptide be considered in all older people with clinical features that suggest type 1 diabetes, with islet autoantibodies being the initial test of choice in short-duration disease (<3 years) and C-peptide the test of choice at longer durations.

Diabetes-Associated Comorbidities and Complications