nature reviews endocrinology https://doi.org/10.

1038/s41574-023-00853-0

Review article Check for updates

Heterogeneity and endotypes

in type 1 diabetes mellitus
Maria J. Redondo 1
& Noel G. Morgan 2

Abstract Sections

Despite major advances over the past decade, prevention and treatment Introduction

of type 1 diabetes mellitus (T1DM) remain suboptimal, with large and Heterogeneity of T1DM
unexplained variations in individual responses to interventions. The T1DM endotypes
current classification schema for diabetes mellitus does not capture
Implications
the complexity of this disease or guide clinical management effectively.
Conclusions
One of the approaches to achieve the goal of applying precision medicine
in diabetes mellitus is to identify endotypes (that is, well-defined
subtypes) of the disease each of which has a distinct aetiopathogenesis
that might be amenable to specific interventions. Here, we describe
epidemiological, clinical, genetic, immunological, histological and
metabolic differences within T1DM that, together, suggest heterogeneity
in its aetiology and pathogenesis. We then present the emerging
endotypes and their impact on T1DM prediction, prevention and
treatment.

Paediatric Diabetes & Endocrinology, Texas Children’s Hospital, Baylor College of Medicine, Houston, TX, USA.
1

Exeter Centre of Excellence for Diabetes Research (EXCEED), Department of Clinical and Biomedical and Science,
2

University of Exeter Medical School, Exeter, UK. e-mail: [email protected]

Nature Reviews Endocrinology

Review article

Key points from birth in the Colorado Diabetes Autoimmune Study in the Young
(DAISY) study, the Finnish Type 1 Diabetes Prediction and Prevention
(DIPP) study and the German BABYDIAB and BABYDIET studies9. The
•• Type 1 diabetes mellitus (T1DM) is heterogeneous; defining lifetime risk of developing clinical T1DM is almost 100% once an indi-
endotypes, or disease subtypes each of which has a unique vidual reaches stage 2 T1DM, which is defined by glucose levels during
aetiopathogenesis that is amenable to a particular intervention, an oral glucose tolerance test that are abnormally elevated, but still
will help apply precision medicine to T1DM. under the diagnostic thresholds for diabetes mellitus, in addition to
multiple autoantibody positivity10.
•• T1DM endotype 1 (T1DE1) includes T1DM diagnosed in early childhood Both genetics and environmental factors contribute to the ini-
and is characterized by extensive, early, β-cell destruction, aggressive tiation of islet autoimmunity and its progression to clinical T1DM.
insulitis with abundant CD8+ T and CD20+ B cells, aberrant proinsulin Half of the genetic risk of developing T1DM is linked to the HLA
processing and an elevated circulating proinsulin to C-peptide ratio. region, particularly class II (HLA-DR, HLA-DQ and HLA-DP) and class I
(HLA-A, HLA-B and HLA-C) genes. The strongest association occurs
•• T1DM endotype 2 (T1DE2) includes T1DM diagnosed in adolescence with the HLA-DR4 allele (OR 6.81) and with the HLA-DR3 allele (OR
or adulthood and is characterized by retention of many residual insulin- 3.54)11. The strongest protective alleles are HLA-DR2 (OR 0.21), HLA-
containing islets and without insulitis, fewer infiltrating CD8+ T cells, DR5 (OR 0.30) and HLA-DR7 (OR 0.24)11. Of note, unlike in people of
few CD20+ B cells, normal proinsulin processing and lower proinsulin European descent, in people of African descent, the DR3 and DR7
to C-peptide ratio than T1DE2. haplotypes confer, respectively, protection and susceptibility, while
in people of Asian descent DRB1*09:01 confers a high risk12. There are
•• Evidence is emerging that T1DE1 might respond better than T1DE2 >75 non-HLA loci associated with T1DM, with most regulating immune
to interventional immunotherapy with agents targeted to specific functions13; however, the majority are also present in β-cells, which rein-
immune cell subsets, such as rituximab or teplizumab, while GAD–alum forces the prominent role of the β-cell in the pathogenesis of T1DM. Out
therapy might be effective for treating T1DE2. of these non-HLA loci, the insulin variable number of tandem repeats
(INS-VNTR) polymorphism found in the INS promoter displays the
•• The T1DE2 endotype could underlie a spectrum of phenotypes strongest association with T1DM and regulates thymic immune toler-
with different degrees of severity of the autoimmune attack and thus, ance to insulin. Other robustly linked genes are CTLA4, PTPN22 and
different rates of progression to insulin dependence, ranging from Il2RA14. However, for many single nucleotide polymorphisms (SNPs)
classic T1DM to latent autoimmune diabetes mellitus in adults or slowly associated with diabetes mellitus, the relevant gene is unknown15.
progressive insulin-dependent diabetes mellitus. Among the various environmental exposures implicated in T1DM,
coxsackievirus infection seems to be strongly influential16 and vari-
•• Whether T1DM endotypes exist is still a matter of debate, but data ations in the microbiome are also increasingly reported17. Similar to
are accumulating that support this framework, which will benefit genetic factors, the environmental factors that promote islet auto-
from further research, including testing interventions directed to their immunity can vary from those that mediate disease progression.
underlying aetiopathogenesis. Moreover, environment and genetics can interact18, complicating
studies on aetiology and prediction. Nevertheless, algorithms that
incorporate demographic, genetic, immunological and metabolic
Introduction factors accurately predict T1DM19,20 and, although most studies still
Type 1 diabetes mellitus (T1DM) is currently diagnosed in individuals focus on genetically predisposed individuals, strategies are being pro-
with insulin deficiency attributed to islet autoimmunity1. Major actors posed to screen the general population for T1DM risk in a cost-effective
in the autoimmune attack against β-cells are antigen-specific autoreac- manner21,22.
tive T cells that are present in serum and islets of affected individuals2, Although early signs of insulin secretory capacity loss (measured
regulatory T cells that fail to control effector cell populations3, and as low C-peptide levels) can often be seen when islet autoantibodies
β-cell abnormalities that support autoimmunity. These β-cell abnor- first appear in serum23, the decline in levels of C-peptide begins to
malities include HLA overexpression4 and increased protein misfolding accelerate around 2 years before diagnosis24. After diagnosis of clinical
leading to neoantigen generation5,6. B cells, neutrophils, macrophages, T1DM, treatment with insulin improves glycaemia, which is often fol-
dendritic cells and natural killer cells have also been implicated in T1DM lowed by a partial recovery of endogenous insulin secretory capacity
pathogenesis7. The involvement of circulating B cells is unsurprising (partial remission, or ‘honeymoon’). However, this partial remission
given that most people with T1DM produce autoantibodies, which are is only transient as the underlying β-cell demise continues. In a long-
secreted by plasma cells expanded from B cell precursors. However, term study, the exponential decline in C-peptide levels was found to
the involvement of B cells within pancreatic islets themselves is less slow down and stabilize about 7 years after the onset of clinical T1DM25.
common in T1DM and is still poorly understood. It has been hypoth- Although most individuals with long-standing T1DM do not produce
esized, however, that these B cells might serve as antigen-presenting clinically significant amounts of insulin, some insulin secretion can
cells operating at the site of tissue inflammation8. persist for several decades26. The decline in levels of C-peptide after
Islet autoantibodies, although not pathogenic, are used as bio- the onset of T1DM is at least partially prevented by treatment with the
markers for T1DM prediction and diagnosis because they are detectable anti-CD3 humanized monoclonal antibody teplizumab27, the anti-CD20
in serum before, at and, often, for years after clinical onset. The 10-year monoclonal antibody rituximab28, the anti-TNF monoclonal antibody
risk of developing clinical (also known as stage 3) T1DM increased from golimumab29, low dose anti-thymocyte globulin (ATG)30,31 and the cal-
15% with a single positive autoantibody to 70% with multiple posi- cium channel blocker verapamil32, among other agents. Teplizumab has
tive autoantibodies (stage 1 T1DM) in an analysis of children enrolled proven to be safe and effective in delaying the progression to clinical

Nature Reviews Endocrinology

Review article

T1DM in autoantibody-positive individuals33,34. However, responses to Islet autoimmunity

teplizumab, rituximab, golimumab, ATG, verapamil and other drugs Heterogeneity in islet autoimmunity is evident by the characteristics
aiming to modify the natural course of T1DM before or after its clinical of both infiltrating immune cells in and around pancreatic islets (Fig. 1)
diagnosis are variable for reasons still poorly understood35. and circulating islet autoantibodies (for example, to GAD65, IA-2 (also
T1DM is a highly heterogeneous disease, as demonstrated by wide known as ICA512), ZnT8 and insulin). A small percentage of individuals
differences in epidemiology, aetiopathogenesis, clinical course and lack measurable serum islet autoantibodies60,61, possibly due to islet
responses to intervention. This heterogeneity poses challenges to dysfunction resulting from cellular autoimmunity without circulating
disease prediction, prevention, diagnosis and treatment. More broadly, humoral markers or islet dysfunction resulting from non-autoimmune
the current classification schema for diabetes mellitus does not capture causes, such as mutations in unidentified genes62,63. Ketosis-prone
its complexity or guide clinical management effectively36,37. One of diabetes mellitus, which is more common in individuals of African or
the approaches to achieve the goal of applying precision medicine in Asian ancestry than in people of European ancestry, displays intermit-
diabetes mellitus (defined as “the right treatment for the right patient tent insulin dependence, and can lack autoantibodies and typical HLA
at the right time”38) involves identifying ‘endotypes’, that is, subtypes associations64.
of disease each of which has a distinct aetiopathogenesis that might Islet autoimmunity-related seroconversion often occurs during
be amenable to specific interventions39,40. In this Review, we consider the first 6 years of life but both the time course and the sequence of
the heterogeneity of T1DM, describe the evidence that supports the appearance in serum of specific autoantibody types are variable21,65,66.
emerging concept of T1DM endotypes and discuss the implications This fact might imply an element of stochasticity in autoantibody gen-
of this concept and ongoing efforts in the field. eration, although more specific aetiologies associated with age-related
endotypes have been proposed51,67. The fact that antibody generation
Heterogeneity of T1DM frequently occurs early in life implies that the triggering events for
Epidemiology and clinical characteristics islet autoimmunity can be active at the youngest ages, possibly during
The global incidence of T1DM is estimated as 15 cases per 100,000 the first 2 years of life even if these do not lead immediately to disease
people and the prevalence as 5.9 cases per 10,000 people (95% CI 0.07– onset. Thus, other parameters must dictate the rate at which β-cell
0.12) but there are large geographical differences between countries41 dysfunction and death ensue (such as which endotype manifests).
that could reflect varying environmental triggers and genetic predispo- Whether those who develop T1DM later in life also have a much earlier
sition42. Clinically, perhaps the most extreme variation in T1DM mani- development of islet autoantibodies (consistent with early initiation of
festation is seen with fulminant T1DM in India and east Asia, which often the autoimmune process) is an area that requires further investigation
presents in association with HLA-DRB1*04:05-DQB1*04:01, without in longitudinal studies.
islet autoantibodies and with global pancreatic inflammation43,44. On
the other hand, the most common form of diabetes mellitus in Japan
a b
is slowly progressive insulin-dependent diabetes mellitus (SPIDDM),
which is an autoimmune form of diabetes mellitus that progresses
at a slower pace than typical T1DM45. Slower loss of β-cell function is
also seen in latent autoimmune diabetes mellitus in adults (LADA)46,
although, unlike SPIDDM, this entity is by definition diagnosed in indi-
viduals >30 years of age. Smaller differences between ethnic groups
include those seen in the USA, where Hispanic or non-Hispanic Black
children develop islet autoantibodies at an older age, have lower risk
and slower rates of progression47 and, at the onset of T1DM, are older,
have higher BMI and higher blood levels of glucose than non-Hispanic
white children48,49. A study published in 2023 showed that certain Afri-
can populations have a form of T1DM in which the autoantibodies that
are usually found in European populations are not present50.
T1DM shows profound differences by age51,52. Age at diagnosis
partly reflects the rate of progression through preclinical stages
DAPI CD45 Insulin Glucagon
of T1DM, which occurs more quickly with earlier initiation of islet
autoimmunity. Furthermore, clinical presentation is more severe, Fig. 1 | Immunological heterogeneity in T1DM. Immunofluorescent images
with greater frequency of diabetic ketoacidosis in children than in of pancreatic samples from two individuals with recent-onset type 1 diabetes
adults53. Genetic predisposition can also differ between childhood- mellitus (T1DM). The islets show distinct patterns of endocrine cells (insulin
onset and adult-onset T1DM54,55. T1DM-associated HLA genotypes in yellow and glucagon in blue) and associated lymphocytic infiltrates (CD45+
cells in red). a, Individual with T1DM diagnosed in adulthood. Retention of
are less frequent and the burden of T1DM-related genetic regions
large numbers of β-cells with minimal lymphocytic infiltration is apparent.
(as measured by T1DM genetic risk scores, a weighted combination of
Most lymphocytes are arrayed peripherally and very few of them have migrated
T1DM-associated SNPs in HLA and non-HLA regions) becomes lower
into the core of the islet. b, Individual with T1DM diagnosed at <2 years of age.
as age at diagnosis increases56,57. Furthermore, single autoantibody In contrast to the sample in panel a, the two islets in this sample are heavily
positivity is more frequent in adult-onset than paediatric T1DM58. infiltrated by lymphocytes. These lymphocytes have breached the islet capsule
Although it is now recognized that more cases are diagnosed during and many are in close proximity to β-cells, consistent with an aggressive
adulthood than childhood, adult-onset T1DM is still under-studied autoimmune attack. These strikingly different patterns are typical of those seen
and, in the clinical setting, it is often misclassified as type 2 diabetes among patients developing diabetes mellitus at different ages. Images courtesy
mellitus (T2DM)57,59. of P. Leete.

Nature Reviews Endocrinology

Review article

Several commonly detected autoantigens (including insulin, IA-2 serum levels of triglycerides and insulin resistance were primary
and ZnT8) are synthesized in coordination with each other and local- features.
ized within insulin secretory granules68, hinting that some facet of β-cell Overall, young age and early seroconversion were associated with
physiology might underpin initial autoimmunity. Anti-insulin antibod- more intense inflammatory responses. This finding aligns, in part,
ies tend to develop during the first 2 years of life and their epitope with previous evidence that early age at T1DM onset is associated with
specificity and binding affinities are markedly heterogeneous69. a strongly pro-inflammatory signature mediated by IFNγ74. By contrast,
A second major β-cell autoantigen, the 65 kDa isoform of glutamate older age at onset, extending into the teenage years, is associated with
decarboxylase (GAD65) bucks the secretory granule protein trend a less intense inflammatory milieu, characterized by a primarily IL-10
since, in β-cells, GAD65 resides in small membrane-enclosed vesi- signature. A 2022 study monitoring longitudinal T cell responses in a
cles68,70. Of GAD65 antibodies, those directed against the central and small cohort of children at high genetic risk of T1DM again revealed
carboxy-terminal regions of the GAD65 protein (known as ‘truncated two profiles75. The first featured an increased IFNγ response when
GAD’ antibodies) are most predictive of progression to clinical T1DM71. T cells were exposed to proinsulin or insulin-derived peptides and
In the TEDDY study, clustering 370 children according to their was detectable up to 6 months prior to T1DM onset. The second was
longitudinal profiles of autoreactivity to insulin, GAD and IA-2 revealed characterized by an enhanced regulatory IL-10 response which, at least
stratification of the likelihood of progression to clinical T1DM72. More during the period of study, occurred in children who did not show
granular analysis revealed that earlier age at seroconversion to each progression to T1DM.
of the autoantibodies was the single most important discriminatory
feature in determining probable progression to disease, irrespective Metabolic factors
of the absolute profiles of autoantibody specificity. A classification Some people with T1DM also display features that are typically associ-
and regression tree approach to stratification of residual levels of ated with T2DM, such as insulin resistance, obesity or specific genetic
C-peptide of young people <20 years old was taken in the German associations58,76,77. This observation is not surprising because T2DM
DiMelli study73, where autoantibody status was considered in parallel is common in the general population, and so it is to be expected that
with age at T1DM onset, glucose control indices and BMI. Ten differ- some people with T1DM would independently develop T2DM. High BMI
ent subgroups emerged, among which, seven were autoantibody- has been shown to accelerate the onset of clinical T1DM76,78,79, and this
positive and which included 1,088 (91.2%) of the 1,192 individuals effect was more pronounced in Hispanic (400% increase in risk) than
studied. Age at diagnosis was a critical factor in providing a basis for in non-Hispanic white children (34% increase) enrolled in the Type 1
stratification. It was concluded that mechanisms marked by eleva- Diabetes TrialNet study47 (Fig. 2). Also in the TrialNet study, overweight
tions in levels of IFNγ, IL-10 and TNF are influential in driving autoim- or obesity, older age and having a single positive autoantibody were
munity in children younger than 8 years of age. Compared with this associated with lower Index60, indicating relatively well preserved lev-
subset of young children, a further subset of older youth was char- els of C-peptide in relation to levels of glucose80,81. Furthermore, using
acterized by a more benign inflammatory milieu, in which elevated area-under-the-curve C-peptide and glucose measurements derived

1.0
Hispanic children with overweight or obesity
Cumulative incidence of type 1 diabetes mellitus

Non-Hispanic white children with overweight or obesity

Lean non-Hispanic white children
Lean Hispanic children

0.4

0.2

0.0
0 1 2 3 4
Time from determination of autoantibody positivity (years)
Number at risk
Lean non-Hispanic white children 1,521 835 561 397 275
Non-Hispanic white children with overweight or obesity 435 222 150 96 68
Lean Hispanic children 150 79 50 33 20
Hispanic children with overweight or obesity 82 47 30 20 12

Fig. 2 | Influence of type 2 diabetes mellitus-related factors on type 1 overweight or obesity increased the risk of type 1 diabetes mellitus by 36% in non-
diabetes mellitus development, and the effect of ethnicity. Data from the Hispanic white children (HR 1.36, P = 0.024) while the risk was almost quadrupled
Type 1 Diabetes TrialNet study of autoantibody-positive relatives of individuals in Hispanic children (HR 3.8, P = 0.0026) after adjustment for confounders.
with type 1 diabetes mellitus. Among children less than 12 years of age, having Figure adapted with permission from ref. 47.

Nature Reviews Endocrinology

Review article

from oral glucose tolerance tests to classify autoantibody-positive Table 1 | Geographical variation of T1DM genetic risk
TrialNet participants into 25 metabolic zones revealed that, for the same
level of glucose, participants with higher levels of C-peptide had higher Region Haplotype Prevalence (%)
insulin resistance, older age and fewer immunological and genetic Africa DR3-DQ2.5 7.2
markers of T1DM than participants with lower levels of C-peptide82. DR4-DQ8.1 4.7
The heterogeneity observed among these 25 metabolic zones also sup-
DR9-DQ9.3 0.0
ports the concept that, even within autoantibody-positive individuals
without obvious T2DM risk factors, there is extensive variability in the DR15-DQ6.1 or DR15-DQ6.2 12.8
degree of β-cell dysfunction and endoplasmic reticulum (ER) stress83. DR13-DQ6.3 or DR13-DQ6.9 5.9
Furthermore, obesity might influence the progression of islet
East Asia DR3-DQ2.5 6.8
autoimmunity through inflammation, ER stress and β-cell apoptosis84.
In T2DM, elevation of the ratio of T helper 17 cells to regulatory T cells85, DR4-DQ8.1 5.7
pro-inflammatory cytokines (such as CCL2 and TNF)86, islet-reactive DR9-DQ9.3 15.5
T cells87 and islet autoantibodies87 have been described. In a 2022 DR15-DQ6.1 or DR15-DQ6.2 10.1
study, cellular islet autoimmunity, humoral islet autoimmunity or both
DR13-DQ6.3 or DR13-DQ6.9 2.7
were observed in, respectively, 41.3%, 13.5% and 5.3% of participants
in the Glycaemia Reduction Approaches in Diabetes: A Comparative Europe DR3-DQ2.5 12.2
Effectiveness Study (GRADE)88. Accordingly, among adolescents and DR4-DQ8.1 9.8
children >9 years of age, without the high risk HLA DR4-DQ8 or DR3-
DR9-DQ9.3 1.0
DQ2 haplotypes, islet autoimmunity was more likely to progress in
those with overweight or obesity than in those without89. DR15-DQ6.1 or DR15-DQ6.2 14.1
DR13-DQ6.3 or DR13-DQ6.9 6.8
Genetics Middle East DR3-DQ2.5 8.9
A higher T1DM genetic risk score (indicating a higher genetic burden for
DR4-DQ8.1 8.3
the disease) correlates with higher T1DM risk and more rapid progres-
sion through the preclinical stages19. The ability of genetics to predict DR9-DQ9.3 0.4
risk of progression to clinical T1DM weakens progressively from the DR15-DQ6.1 or DR15-DQ6.2 9.1
initiation of islet autoimmunity with a single autoantibody to the transi- DR13-DQ6.3 or DR13-DQ6.9 5.3
tion to multiple positivity and then to T1DM diagnosis19, and the specific
South Asia DR3-DQ2.5 7.5
genetic regions involved in each step vary90. Yet, T1DM-associated
HLA genotypes were less frequent in individuals with multiple positive DR4-DQ8.1 8.9
autoantibodies who did not progress to T1DM within 10 years after DR9-DQ9.3 0.6
seroconversion as compared with children who presented with clinical
DR15-DQ6.1 or DR15-DQ6.2 19.1
T1DM at <5 years of age91. Moreover, geographical differences can also
be explained, at least in part, by differences in HLA alleles and haplo- DR13-DQ6.3 or DR13-DQ6.9 8.3
types among populations42 (Table 1). Although most genetic studies of The prevalence of five HLA-DR-DQ haplotypes that are associated with susceptibility
or resistance to type 1 diabetes mellitus (T1DM) varies across geographical regions
T1DM have been conducted in individuals of European ancestry, emerg-
(data from ref. 42).
ing cohorts from other ethnicities are revealing variation between eth-
nic groups in the genetic risk conferred by HLA alleles12. As additional
instances of genetic variation underlying heterogeneity in T1DM, DR3 after adjustment for age at onset, diabetes mellitus duration, BMI
and DR4 alleles at the HLA class II locus are differentially associated z-score, sex and African American race95. In addition, participants in the
with both age at onset and first autoantibody generation. Children in T1DM Exchange with established T1DM carrying the T2DM-associated
the TEDDY study who carried HLA-DR3-DQ2 haplotypes were older TCF7L2 SNP, compared with those who did not carry it, were less likely
at diagnosis and were more likely to feature primary autoreactivity to have T1D-associated HLA genotypes96.
against GAD65, whereas those carrying HLA-DR4-DQ8 were diagnosed
earlier in childhood and were more likely to display autoreactivity T1DM endotypes
against insulin as an initial response92. Disease endotypes are defined as having intrinsically unique pathologi-
Further supporting the concept that pathogenic mechanisms cal processes that necessitate specific treatment approaches and have
typically associated with T1DM or T2DM can combine and interact in prognostic implications39. Endotypes differ from phenotypes, which
the same individual, it has been shown that TCF7L2 genetic variants that represent observable characteristics or traits of a disease that do not
confer risk of T2DM also modify the natural course93 and presentation always entail a distinct mechanism. For example, different degrees of
of T1DM77,94. Individuals with new-onset T1DM and a single positive severity or rates of progression of a disease are phenotypic features
autoantibody (that is, only mild autoimmunity) are more likely to have that do not necessarily imply an idiosyncratic pathogenesis. Similarly,
insulin resistance and, among adolescents and adults, carry the T2DM- age or sex of the patient might modulate the expression of a phenotype
associated allele in the TCF7L2 SNP than those with multiple autoanti- without the result qualifying as an endotype.
body positivity77. Consistent with these findings, among donors from On the other hand, a distinct phenotype is often the first indicator
the Network for Pancreatic Organ Donors with Diabetes (nPOD) with of a different pathogenic mechanism and, when a distinctive molecu-
T1DM, those with the TCF7L2 risk allele, compared with donors without lar or cellular mechanism can be attributed, and/or effectiveness of
the allele, had a higher frequency of residual insulin-containing islets a specific treatment proven, the phenotype is better identified as an

Nature Reviews Endocrinology

Review article

endotype or a separate disease. For example, maturity onset of diabetes availability of pancreas recovered from people (especially young chil-
in the young (MODY) was recognized as a separate entity from T1DM or dren) with recent onset disease is severely limited105. The most extensive
T2DM after observing the distinct phenotype of young people with mild collection recovered from children <10 years of age at T1DM onset was
diabetes mellitus and a family history of diabetes mellitus consistent compiled within the UK by Foulis and colleagues106 and is now curated
with Mendelian inheritance97. This phenotype was shown to respond as part of the larger Exeter Archival Diabetes Biobank107. The Network
to non-insulin treatments98,99 and decades later was attributed to spe- for Pancreatic Organ Donors with Diabetes (nPOD), supported by the
cific genetic mutations100. Further research led to the discovery of Juvenile Diabetes Research Foundation, also holds an extensive con-
multiple unique genetic mutations causing specific molecular defects temporary collection108,109, with most donors of pancreatic tissue with
that respond to precise therapeutic approaches, such as a particular recent-onset T1DM being >10 years of age. Study of these samples,
drug or absence of medical treatment101. The strategy of aiming to together with specimens available from a Belgian collection110,111 has
develop disease taxonomy based on discrete biological signatures revealed marked heterogeneity of T1DM on multiple levels.
(endotypes)102 has proven useful to advance targeted therapies in the The first level of heterogeneity lies in the realization that islets
field of asthma and is also being tested to dissect rheumatological within a given pancreas are subject to immune-mediated attack at vari-
diseases103. In summary, phenotypic heterogeneity can support patho- able rates over time112–114. Examination of the tissue reveals distinct foci
genic heterogeneity but proving the existence of separate endotypes of β-cell destruction with islets in some regions apparently untouched
requires the identification of distinct pathogenic mechanisms that are while others, often in close proximity, are devoid of insulin immuno-
amenable to specific treatment. positivity. Again, age is important since children <10 years of age at
disease onset display the least heterogeneity, with the majority of their
T1DM endotypes 1 and 2 β-cells destroyed and most residual insulin-containing islets under
Arguably the most substantial impediment to an improved understand- active inflammatory assault115. By contrast, those who develop T1DM
ing of the heterogeneity of T1DM is the inability to monitor the disease in their teenage years display a much less aggressive disease profile
process in humans in real time at the site of tissue damage. Currently, with many insulin-containing islets retained (frequently >50%), most
studies rely mainly on pancreatic tissue from people with T1DM recov- of which lack inflammatory infiltrates. It has proven difficult to assign
ered after death either at the time of organ donation or at autopsy, these differences to any underlying genetic architecture, but hints
although six pancreas biopsy samples from living patients newly diag- have emerged that certain predisposing SNPs in genes including IKZF3
nosed with T1DM have also been highly informative104. Mercifully, few and Il10 are associated with children diagnosed at <7 years of age116.
individuals now die close to the diagnosis of T1DM; consequently, the Aligned with a variability in the proportions of inflamed islets are
substantial variations in the magnitude and composition of the infil-
trating immune cells. In the very youngest children (<7 years of age),
Table 2 | Characteristics of T1DE1 and T1DE2 the islet-associated inflammatory infiltrates comprise large numbers of
both CD8+ T cells and CD20+ B cells, whereas in older children (≥13 years
Endotype T1DE1 T1DE2
of age) the absolute number of infiltrating CD8+ T cells is typically
Primary genetic HLA DR4-DQ8 HLA DR3-DQ2 much lower and very few CD20+ B cells are detected8,115,117. Importantly,
association these two immunological profiles segregate with age at diagnosis, and
Islet autoantibodies IAA first (<2 years GAD autoantibodies first, they do not represent a continuum. Inevitably, however, although the
old), then IA-2 with then with or without other age dependence of the immune profiles is strict for those <7 years or
or without GAD65 autoantibodies >12 years of age, there is overlap within the intermediate (8–12 years)
autoantibodies
age group. As a result, both profiles can be found among children in
High percentage this intermediate age group. Nevertheless, the two profiles remain fully
IAA+ and IAA titres segregated in that all pancreata examined from children 8–12 years of
at T1DM onset
age display either one pancreatic immunological profile or the other.
Other immunological High percentage Low percentage of CD8+ This finding strongly supports the proposition that the two immune
findings of islet-infiltrating T cells compared with cell profiles reflect differences in disease aetiology. Accordingly, two
CD8+ T cells, high T1DE1, very low percentage
percentage of CD20+ of CD20+ B cells compared immunological endotypes have been proposed: T1DM endotype 1
B cells with T1DE1 (T1DE1) and T1DM endotype 2 (T1DE2)118 (Table 2). Despite this evi-
dence, the concept of endotypes in T1DM remains controversial and,
β-Cell abnormalities High proinsulin to Higher percentage of
C-peptide ratio insulin-containing islets in this Review, we strive to present both perspectives.
compared with T1DE1 Additional histological analysis of pancreatic tissue from people
Low percentage of with T1DM has revealed marked variation in the ability of residual
insulin-containing
cells
β-cells to process insulin correctly between the two proposed endo-
types. In children with T1DM defined as T1DE1, the majority of islets
Abnormal β-cell display evidence of aberrant proinsulin processing, leading to a marked
maturation
increase in the circulating ratio of proinsulin to C-peptide. By contrast,
Response to Yes Less responsive than T1DE1 in children with T1DM defined as T1DE2, most islets retain apparently
immunomodulators normal proinsulin processing and the circulating ratio of proinsulin
Other disease Coeliac disease Thyroid autoimmunity to C-peptide is correspondingly lower118. Measurement of this ratio
associations offers a potential means to differentiate between T1DE1 and T1DE2
IAA, insulin autoantibodies; T1DE1, T1DM endotype 1; T1DE2, T1DM endotype 2; T1DM, type 1 among children who develop T1DM between 8 years and 12 years of
diabetes mellitus. age, where either pancreatic endotype can be found (Fig. 3). Of course,

Nature Reviews Endocrinology

Review article

increases in circulating levels of proinsulin are not confined solely to T1DM onset
T1DM, since increased levels of proinsulin can also occur in T2DM (and <7 years old 7–13 years old 13–18 years old Adult
in older people with T1DM), but both our own analysis118 and those
of others119,120 indicate that the elevation in proinsulin to C-peptide
ratio is particularly enhanced in young children with T1DM (classified
as T1DE1). This observation is consistent with histological evidence
implying that the process of proinsulin processing is affected to a much
greater extent in these patients.
When analysing proinsulin processing in children with T1DM ulti-
mately defined as T1DE2, a particularly important finding is that these
children have two different populations of pancreatic islets118. The first
of these populations has apparently normal segregation of insulin and
proinsulin within β-cells while the second population displays aberrant
insulin processing equivalent to that seen in T1DE1. More strikingly still, T1DE1 T1DE2 Type 2 diabetes mellitus Double diabetes mellitus

in pancreas samples from patients with longer duration T1DM, this lat- Fig. 3 | Conceptual model for the distribution of type 1 diabetes mellitus
ter population of islets was found to be missing, suggesting that these endotypes by age at onset. Type 1 diabetes mellitus (T1DM) can develop as
islets might have been targeted selectively during the autoimmune either T1DM endotype 1 (T1DE1, represented in red) or endotype 2 (T1DE2,
attack. The residual insulin-containing islets in these patients (which represented in yellow). T1DE1 is the predominant endotype in children diagnosed
represented a much higher proportion of the total number of islets with T1DM less than 7 years of age and its prevalence decreases sharply in
than in T1DE1, as judged by co-immunostaining with an anti-glucagon individuals diagnosed at older ages. T1DE2 is the predominant type after age 7
antibody) did not display aberrant proinsulin processing118. Thus, dif- years of age and increases with age, becoming the predominant T1DM endotype
ferences in immune cell profiles, proinsulin processing, the proportion after 13 years of age. In the intermediate age group (7–13 years of age), some
children have T1DE2 and other children have T1DE2. Type 2 diabetes mellitus,
of residual insulin-containing islets and the extent of β-cell loss all differ
which is highly prevalent in adults but also appears in younger individuals, can
between T1DE1 and T1DE2 (Table 2).
coexist with T1DE2 and modify its features, resulting in double diabetes mellitus;
It must be acknowledged that many of these initial studies were
this interaction is represented here as blue and yellow stripes.
undertaken by sampling only a proportion (sometimes a small propor-
tion) of the islets present in each pancreas section. This limitation was
an inevitable consequence of the time-consuming nature of manual
assessment. To ensure that the conclusions are fully representative of between cases represents a continuum of heterogeneity rather than
the wider islet population and with the advent of whole section scan- pointing to distinct endotypes. We do not subscribe to these views
ning and software-based analysis, the data have now been expanded and would emphasize the absolute separation of the proposed endo-
and identical outcomes obtained from many thousands of islets rather types, T1DE1 and T1DE2, among children <8 years of age versus those
than a few tens to hundreds, as had been studied previously (N.G.M., >13 years of age. Even if this current differentiation turns out to be an
unpublished work). over-simplification, this does not exclude our firm conclusion that
Work from 2022 has revealed that islets of the youngest children, the pancreatic immune profiles define different disease aetiologies.
<4 years of age, progressing to T1DM can selectively express a variant Rather, we emphasize the adage that ‘the exception proves (tests) the
of neuropilin 1, which renders their β-cells refractory to VEGF signal- rule’ but, so far, no exceptions have been found.
ling and thereby impairs islet maturation121. By contrast, the β-cells of
children who are older at onset can express a VEGF-responsive isoform Non-classic forms of T1DM with slow decline of β-cell function
of neuropilin 1, and this difference has been proposed as a potential Although T1DE2, the endotype observed in adolescent-onset and adult-
molecular basis for the development of T1DM endotypes in children121. onset T1DM, has a slower progression of loss of β-cell function (both
This hypothesis remains to be verified and it would be of immediate before and after diagnosis) than that in very young children with T1DE1
interest to establish whether the islets of children with T1DE1 express (Fig. 4a,b), the severity and rate of autoimmune islet destruction vary
the variant form of neuropilin 1 preferentially. among individuals with T1DE2. Many of these individuals develop T1DM
In summarizing the histological evidence that has spawned the with classic features (that is, rapid loss of β-cell function), probably cor-
endotype concept in T1DM, it must be accepted that the existence responding to the cluster identified as ‘severe autoimmune diabetes’
of endotypes has been concluded based on a relatively small number of (SAID) by Ahlqvist et al.122. At the other extreme of the spectrum are
cases. However, there is good reason for this fact because few pancreas those with SPIDDM, which has been well described in the Japanese popu-
samples taken from people with recent-onset T1DM exist worldwide. lation123, and LADA46, which is, by definition, diagnosed in individuals
Moreover, many of these samples come from people with older-onset above 30 years of age.
disease, and few researchers have had the opportunity to study recent- We propose that SPIDDM and LADA are extreme cases of T1DE2,
onset disease in the pancreas of young children. This lack of available with weaker islet autoimmunity than seen in typical T1DE2. This con-
pancreatic samples can be a cause of scepticism of the endotype con- cept of weaker islet autoimmunity is supported by the observation
cept as few histopathologists have seen at first hand the differences that individuals with LADA or SPIDDM123 often develop positivity to
in immune profile between T1DE1 and T1DE2. Inevitably, this fact can only one islet autoantigen, most commonly GADA. Compared with the
lead to questioning of the import of any differences reported and to aggressive attack on β-cells that occurs in classic T1DM, the relatively
the notion that they are likely to reflect changes in the intensity of the mild islet autoimmunity characteristic of LADA and SPIDDM takes a
immune attack rather than materially different underlying mecha- longer time to destroy β-cells to the point of causing clinical diabetes
nisms. It is also the case that some could consider that the variation mellitus. Patients with LADA or SPIDDM initially have sufficient β-cell

Nature Reviews Endocrinology

Review article

a T1DE1 b T1DE2

Insulin secretion capacity (%)

Insulin secretion capacity (%)

Early childhood Adolescence or adulthood
Age (years) Age (years)

c LADA and SPIDDM: special cases of T1DE2 d T1DE2 with coexistence of T2DM mechanisms
(double diabetes mellitus)
Insulin secretion capacity (%)

Insulin secretion capacity (%)

Adolescence or Adolescence or adulthood

adulthood
Age (years) Age (years)

Threshold for diabetes mellitus Threshold for diabetes mellitus with insulin
resistance (double diabetes mellitus)

Fig. 4 | Conceptual model for the variability in trajectory of insulin secretory insulin needs and production, which causes diabetes mellitus. Therefore, the
capacity using the T1DM endotype framework. a, In type 1 diabetes mellitus combination of T1DM and type 2 diabetes mellitus (T2DM) risk factors (that
(T1DM) endotype 1 (T1DE1), β-cell function declines rapidly and clinical diabetes is, double diabetes mellitus) is common in individuals with slowly progressive
mellitus develops in early childhood. b, In T1DM endotype 2 (T1DE2), β-cell islet autoimmunity who show progression to diabetes mellitus. In LADA and
function declines more slowly and clinical diabetes mellitus develops later SPIDDM, insulin resistance is often, but not always (represented by the dashed
in life compared with T1DE1. c, Latent autoimmune diabetes of adults (LADA) line). d, T2DM risk factors can coexist with islet autoimmunity, accelerating
and slowly progressive T1DM (SPIDDM) are special cases of T1DE2 with slower the progression to clinical diabetes mellitus through insulin resistance,
loss of β-cell function than classic T1DM. Thus, progression to clinical diabetes inflammation, endoplasmic reticulum stress and other factors. Because
mellitus, if it happens, occurs late in life, such as in adolescence or adult life. the action of T2DM risk factors is fairly slow, their effect on the preclinical
If present, additional diabetogenic factors (insulin resistance, non-autoimmune progression to diabetes mellitus is more appreciable in individuals with slower
insulin secretion defects) accelerate the development of an imbalance between β-cell function decline than in those with rapid loss of β-cells.

function to maintain insulin independence even after diagnosis, but The phenotype that results from the combination of T1DM and
this phase is only transient as the disease progresses and more β-cells T2DM pathogenic mechanisms has been given different names in
are lost (Fig. 5). the literature, including double diabetes mellitus (DDM) and type
In classic T1DM, progression to clinical diabetes mellitus is faster 1.5 diabetes mellitus124 (see Box 1). Persistent residual β-cell function,
than in SPIDDM and LADA, and the thresholds for clinical diabetes although not sustained, underlies the response of LADA to non-insulin
mellitus and insulin dependence are crossed almost simultaneously. therapies such as certain thiazolidinediones125, dipeptidylpeptidase IV
SPIDDM and LADA have strong commonalities with classic T1DM, inhibitors126, disease-modifying therapies such as alum-formulated
including similar HLA and non-HLA genetic regions, positive islet recombinant GAD65 (GAD–alum)127, and other agents128, but their
autoantibodies, increased personal and family history of other autoim- respective efficacies require more detailed verification.
mune conditions, and decline in β-cell function. However, the charac- It has been proposed that LADA represents a further T1DM endo-
teristically slow decline in β-cell function in SPIDDM and LADA leads to type129. However, in our view, there is no evidence to date that the
older age at diabetes mellitus onset or even the absence of progression pathogenic mechanisms that lead to β-cell function loss could be
to clinical diabetes mellitus except in the presence of additional dia- different in LADA or SPIDDM than in T1DE2. Instead, these forms,
betogenic factors (Fig. 4c). These additional diabetogenic causes are pheno­typically different from classic T1DM, can be explained by a lower
most often T2DM-related factors, which are collectively very prevalent degree of severity and a slower rate of disease progression compared
in the general population. Indeed, although the genetic architecture with T1DE2. Adding to the ongoing controversy, other investigators
of LADA is closest to that of T1DM, it has some T2DM burden55 such as argue that LADA represents a mix of two subsets of individuals with
TCF7L2 SNPs or the T2DM associated HNF1A locus. either T1DM or T2DM130.

Nature Reviews Endocrinology

Review article

T2DM pathogenic mechanisms in individuals with T1DE2 Among these is the observation that the early development (<2 years
Insulin resistance, obesity or specific T2DM genetic associa- of age) of autoantibodies against insulin and IA-2 that then persist is
tions58,76,77 can accompany not only LADA or SPIDDM but also classic highly correlated with the subsequent onset of diabetes mellitus within
T1DM58,76,77,80,89,94,131. As T2DM rarely develops in prepubertal children132, 5 years (that is, by 7 years of age)72. This would suggest that these chil-
it most frequently coexists, in the preclinical stages, with T1DE2, which dren correspond to those defined by the pancreatic endotype T1DE1.
develops after age 7 years, as opposed to T1DE1, which is seen in chil- Children in whom IA-2 antibodies did not persist were fewer in number
dren younger than 13 years of age (Figs. 4c,d;5). As reviewed in the and it seems premature to assign these to a specific subgroup. Never-
section ‘Heterogeneity of T1DM’, insulin resistance, obesity and genes theless, it is of interest to note that these children developed symptoms
that are typically associated with T2DM can modify the progression of of T1DM slightly later than children with other autoantibody patterns,
islet autoimmunity, accelerate the diagnosis of T1DM and modify its indicating that they could be included among children in whom the islet
clinical presentation and course131. Furthermore, insulin resistance is immunological profile is categorized as T1DE1 but who develop the
an underlying pathogenic feature that can be targeted for treatment. symptoms of T1DM in the intermediate age range between 8 years and
Indeed, metformin, as an adjunct to insulin therapy, improves insulin 12 years of age. Further studies are required to substantiate this notion.
resistance in adolescents with T1DM and obesity133,134. Moreover, the A group of children in whom stable anti-GAD antibodies persisted in the
influence of obesity on the initiation and progression of autoimmunity89 absence of antibodies to IA-2, developed symptoms of T1DM at still later
suggests that obesity-related mechanisms might be targeted for pre- ages during childhood and thus might be considered to have T1DE2.
vention of diabetes mellitus in autoantibody-positive persons. Given Superficially, these findings seem to provide a cohesive set of rela-
that endotypes are defined as disease subtypes with unique pathogenic tionships, where autoantibody types are differentially associated with
mechanisms that could warrant specific preventive and therapeu- clusters of HLA haplotypes, histopathological findings, T1DM risk and
tic strategies, DDM might qualify as a T1DM endotype. On the other ages at presentation, but it is also clear that many children lie outside
hand, as opposed to the clear distinction between T1DE1 and T1DE2 the assigned autoantibody groups. We conclude, therefore, that the
at the individual level, T2DM-related processes (such as insulin resist- definition of pancreatic endotypes according to autoantibody status is
ance) can combine with and influence islet autoimmunity. In addition, not (yet) a reliable index. This caveat might seem disappointing but, in
T2DM risk factors are continuous variables without sharp thresholds our view, it should not deter those working in the field from undertak-
and, therefore, using them as criteria for rigid classification will con- ing additional work designed to verify (or refute) the endotype concept.
tinue to leave heterogeneity within and overlaps between categories36. The ‘Holy Grail’ of the endotype proposal does not lie in the achieve-
An alternative approach is to avoid further categorization of diabetes ment of an ever more complex system of disease classification but,
mellitus and simply evaluate, in each individual with T1DM, whether the rather, it seeks to facilitate the design of (immuno)therapies that are
presence and degree of T2DM mechanisms warrant specific treatment tailored mechanistically to the precise disease aetiology in each person.
or preventive measures. It has been suggested that the finding that T1DM disease proceeds
more quickly in children less than 7 years of age than in those who are
Implications older does not necessarily imply that different therapeutic approaches
The findings we summarize here have potentially important impli- are warranted for these two groups. Researchers have expressed con-
cations for interventional immunotherapy, as illustrated when con- cerns that further subdivision of an already modestly remunerative
sidering the efficacy of the anti-CD20 monoclonal antibody reagent, disease area (T1DM) could lead to reduced therapeutic investment,
rituximab28,135. Clinical trial data hint that a delay in disease onset is
achieved most readily in young children, as might be predicted if CD20+ LADA and SPIDDM Threshold for diabetes mellitus
B cells have a more critical role in driving β-cell destruction at younger Special cases of T1DE2 (still insulin-independent)
ages, as seen in T1DE1. Similar considerations can also be extended to Threshold for insulin dependence
the anti-CD3 humanized monoclonal antibody teplizumab34,136, which
Insulin secretion capacity (%)

might be predicted to be particularly effective in patients with the

most intense islet inflammation that is also a characteristic of T1DE1.
Based on histologically defined endotypes, individuals who respond
to teplizumab would be those with the greatest elevation in circulating
proinsulin to C-peptide ratios. Consistently, response to teplizumab for
prevention of progression from stage 2 to stage 3 T1DM is associated
with carrying HLA-DR4-DQ8 (more common in people with T1DE1)
and lacking HLA-DR3-DQ2 (this haplotype is more common in people
with T1DE2)33. By contrast, GAD–alum therapy might be more effec-
tive in patients with HLA DR3-DQ2 (ref. 137). Unfortunately, the initial
results that higher insulin autoantibody titres might predict response Age (years)
to oral insulin138 were not confirmed by a follow-up trial139, while post
Fig. 5 | Conceptual model for the trajectory of insulin secretory capacity in
hoc analyses of the follow-up trial138 suggested that elevated Diabetes
slowly progressing forms of type 1 diabetes mellitus. In contrast to classic
Prevention Trial Risk Score (DPTRS) (a metabolic score) was associated type 1 diabetes mellitus, the rate of β-cell function decline in latent autoimmune
with response to oral insulin therapy140. diabetes of adults (LADA) and slowly progressive type 1 diabetes mellitus
Clearly, an important goal is to link pancreatic endotypes with (SPIDDM) is lower. This explains the long time that elapses between crossing
readily measured parameters in the circulation and, as noted, the the threshold of diabetes mellitus with sufficient insulin secretion to not require
proinsulin to C-peptide ratio is one such measure. Autoantibody pro- exogenous insulin (pink line), and insulin secretion falling under the threshold
files might be another, and some important trends are emerging. for insulin dependence (red line).

Nature Reviews Endocrinology

Review article

Box 1

Non-classic forms of diabetes mellitus with T1DM and T2DM

characteristics
Latent autoimmune diabetes of adults (LADA) Double diabetes mellitus (DDM)
T1DM resulting from a slowly progressive attack on β-cells that, to Diabetes mellitus with elements of both T1DM (usually islet
reach the threshold for diabetes mellitus, requires more time than autoimmunity) and T2DM (for example, insulin resistance). When
classic type 1 diabetes mellitus (T1DM) (by definition, LADA appears present during the preclinical stages of T1DM, T2DM risk factors can
in people over 30 years of age), frequently in addition to other influence the risk and rate progression of islet autoimmunity and
diabetogenic factors (such as insulin resistance and/or insufficient the development of clinical (stage 3) T1DM. In patients with slowly
β-cell function). Patients cross the threshold for diabetes mellitus but progressive islet autoimmunity (thus, at risk of LADA or SPIDDM),
are still insulin-independent (as in type 2 diabetes mellitus (T2DM)) T2DM can be a definitive factor in the progression to diabetes
for at least 6 months (by definition) until further loss of β-cells causes mellitus143. Whether DDM constitutes an endotype is unclear.
insulin dependence (as in classic T1DM).
Type 1.5 diabetes mellitus
Slowly progressive insulin dependent diabetes mellitus This term is often applied in two settings: first, as a synonym of
(SPIDDM) DDM: diabetes mellitus with elements of both T1DM (usually islet
Caused by slowly progressive autoimmune destruction of β-cells autoimmunity) and T2DM (for example, insulin resistance); and
and, therefore, progression to clinical diabetes mellitus takes longer second, slowly progressive T1DM that initially presents as non-
than in typical T1DM, as in LADA. Clinical diabetes mellitus develops insulin-dependent (frequently leading to a misdiagnosis of T2DM,
at an older age than in typical T1DM and often does so with the aid of particularly if islet autoantibodies are not measured) but progresses
additional T2DM-related mechanisms. Unlike LADA, the definition to insulin dependence faster than expected for T2DM, which often
of SPIDDM is not confined to individuals >30 years who had at least prompts testing that reveals autoantibody positivity and a revision
6 months of insulin independence. of the diagnosis as T1DM.

particularly from pharmaceutical companies. However, by predicting We describe the rationale for the identification of endotypes;
responders to a specific agent, endotypes would improve the risk to that is, pathogenically unique disease subtypes that require specific
benefit ratio of disease-modifying therapies and decrease the number treatment. We present our views, based on our interpretation of the
needed to treat141, a concept proven useful in assisting decision making. data available to date, on the conceivable existence of two immuno-
There is also a question about the possibility that age itself is an logical endotypes, T1DE1 and T1DE2. T1DE1, which appears in children
important variable and that islet autoimmunity develops in children less than 13 years of age, is characterized by near-total loss of β-cells,
at a time when their immune system (and their pancreatic islets) is profuse inflammatory infiltrate of CD8+ T cells and CD20+ B cells and
undergoing radical change. In the view of some, these caveats are aberrant proinsulin processing leading to an elevated proinsulin to
sufficiently important that the notion that endotypes of T1DM exist C-peptide ratio. All of these factors are markedly milder in T1DE2,
is an obfuscation driven by underlying tissue remodelling. We do not which is observed after 7 years of age. Children between the ages of
favour this view but would argue that it is incumbent on those work- 7 years and 13 years seem to develop one or the other endotype. Under
ing in the field to press on with the important task of defining disease this framework, SPIDDM and LADA can be considered special cases
aetiologies in T1DM. Only when these are understood more fully, will of the T1DE2 endotype with particularly protracted disease courses.
the therapeutic relevance of the endotype concept be established. Moreover, while it can be argued that the disease that results from the
combination of T1DM and T2DM (DDM) meets the definition of an
Conclusions endotype, this categorization is hindered by the lack of rigid cut-offs
From within the wide heterogeneity of T1DM, endotypes are emerg- to define the presence or absence of T2DM factors.
ing that can be used to improve prediction of T1DM development and The implications of T1DE1 and T1DE2 on prevention and treatment
response to preventive and therapeutic interventions. In this Review, are beginning to be observed in the differential response to disease-
we present evidence of variability within T1DM in aspects ranging from modifying therapies, such as rituximab, teplizumab or GAD–alum. It
epidemiology to clinical presentations, such as paediatric, adult-onset, must be noted that there are detractors of the concept of endotypes
slowly progressing forms (SPIDDM or LADA) or fulminant onset. Hetero­ and that, overall, there is agreement that more data are needed to
geneous histopathological, immunological and genetic features cluster address the outstanding questions.
into two major subgroups that, to date, are best defined on the basis The Precision Medicine in Diabetes Initiative supported by the
of the age when islet autoantibodies become measurable or clinical American Diabetes Association and the European Association for
T1DM develops. In addition, T2DM-related pathological processes that the Study of Diabetes has engaged a large number of international
can develop in adolescents and adults influence T1DM progression experts in diabetes mellitus to conduct an extensive, ongoing review of
and presentation. the literature on precision medicine in prediction, prevention, diagnosis

Nature Reviews Endocrinology

Review article

and prevention in T1DM and other diabetes mellitus types. With most 28. Pescovitz, M. D. et al. Rituximab, B-lymphocyte depletion, and preservation of beta-cell
function. N. Engl. J. Med. 361, 2143–2152 (2009).
of the knowledge on T1DM heterogeneity stemming from post hoc and 29. Quattrin, T. et al. Golimumab and beta-cell function in youth with new-onset type 1
secondary analyses, validation of findings in adequately powered pro- diabetes. N. Engl. J. Med. 383, 2007–2017 (2020).
spective studies and in clinical trials is among the most pressing issues. 30. Lin, A. et al. Low-dose ATG/GCSF in established type 1 diabetes: a five-year follow-up
report. Diabetes 70, 1123–1129 (2021).
In addition, before the concept of endotypes can be used in clinical 31. Haller, M. J. et al. Low-dose anti-thymocyte globulin (ATG) preserves β-cell function
practice, biomarkers that identify the driving pathogenic process and and improves HbA1c in new-onset type 1 diabetes. Diabetes Care 41, 1917–1925 (2018).
monitor response to treatment must be developed and validated142. 32. Forlenza, G. P. et al. Effect of verapamil on pancreatic beta cell function in newly
diagnosed pediatric type 1 diabetes: a randomized clinical trial. J. Am. Med. Assoc. 329,
It is expected that these and other efforts will bring clarity to the con- 990–999 (2023).
cept of T1DM endotypes and facilitate translation and implementation 33. Herold, K. C. et al. An anti-CD3 antibody, teplizumab, in relatives at risk for type 1
in clinical and research practices. diabetes. N. Engl. J. Med. 381, 603–613 (2019).
34. Sims, E. K. et al. Teplizumab improves and stabilizes beta cell function in
antibody-positive high-risk individuals. Sci. Transl. Med. 13, eabc8980 (2021).
Published online: xx xx xxxx 35. Greenbaum, C., VanBuecken, D. & Lord, S. Disease-modifying therapies in type 1
diabetes: a look into the future of diabetes practice. Drugs 79, 43–61 (2019).
References 36. Redondo, M. J. et al. The clinical consequences of heterogeneity within and between
1. American Diabetes Association Professional Practice Committee. 2. Classification and different diabetes types. Diabetologia 63, 2040–2048 (2020).
diagnosis of diabetes: standards of medical care in diabetes–2022. Diabetes care 45, 37. Siller, A. F. et al. Challenges in the diagnosis of diabetes type in pediatrics.
S17–S38 (2022). Pediatr. Diabetes 21, 1064–1073 (2020).
2. Babon, J. A. et al. Analysis of self-antigen specificity of islet-infiltrating T cells from 38. Cefalu, W. T. et al. Heterogeneity of diabetes: β-cells, phenotypes, and precision
human donors with type 1 diabetes. Nat. Med. 22, 1482–1487 (2016). medicine: proceedings of an international symposium of the Canadian Institutes of
3. Lindley, S. et al. Defective suppressor function in CD4+CD25+ T-cells from patients with Health Research’s Institute of Nutrition, Metabolism and Diabetes and the U.S. National
type 1 diabetes. Diabetes 54, 92–99 (2005). Institutes of Health’s National Institute of Diabetes and Digestive and Kidney Diseases.
4. Benkahla, M. A. et al. HLA class I hyper-expression unmasks beta cells but not alpha cells Diabetes Care 45, 3–22 (2022).
to the immune system in pre-diabetes. J. Autoimmun. 119, 102628 (2021). 39. Battaglia, M. et al. Introducing the endotype concept to address the challenge of disease
5. Marhfour, I. et al. Expression of endoplasmic reticulum stress markers in the islets heterogeneity in type 1 diabetes. Diabetes Care 43, 5–12 (2020).
of patients with type 1 diabetes. Diabetologia 55, 2417–2420 (2012). 40. Deligne, C., You, S. & Mallone, R. Personalized immunotherapies for type 1 diabetes:
6. Mallone, R., Halliez, C., Rui, J. & Herold, K. C. The β-cell in type 1 diabetes pathogenesis: who, what, when, and how? J. Pers. Med. 12, 542 (2022).
a victim of circumstances or an instigator of tragic events? Diabetes 71, 1603–1610 (2022). 41. Mobasseri, M. et al. Prevalence and incidence of type 1 diabetes in the world:
7. Khosravi-Maharlooei, M. et al. Modeling human T1D-associated autoimmune processes. a systematic review and meta-analysis. Health Promot. Perspect. 10, 98–115 (2020).
Mol. Metab. 56, 101417 (2022). 42. Redondo, M. J. et al. Type 1 diabetes in diverse ancestries and the use of genetic risk
8. Leete, P. & Morgan, N. G. Footprints of immune cells in the pancreas in type 1 diabetes; scores. Lancet Diabetes Endocrinol. 10, 597–608 (2022).
to “B” or not to “B”: is that still the question? Front. Endocrinol. 12, 617437 (2021). 43. Wang, Z. et al. Immunological aspects of fulminant type 1 diabetes in Chinese.
9. Ziegler, A. G. et al. Seroconversion to multiple islet autoantibodies and risk of J. Immunol. Res. 2016, 1858202 (2016).
progression to diabetes in children. J. Am. Med. Assoc. 309, 2473–2479 (2013). 44. Kawabata, Y. et al. Differential association of HLA with three subtypes of type 1 diabetes:
10. Krischer, J. P. Type 1 Diabetes TrialNet Study Group. The use of intermediate endpoints fulminant, slowly progressive and acute-onset. Diabetologia 52, 2513–2521 (2009).
in the design of type 1 diabetes prevention trials. Diabetologia 56, 1919–1924 (2013). 45. Kobayashi, T. Subtype of insulin-dependent diabetes mellitus (IDDM) in Japan: slowly
11. Rich, S. S., Weitkamp, L. R. & Barbosa, J. Genetic heterogeneity of insulin-dependent progressive IDDM – the clinical characteristics and pathogenesis of the syndrome.
(type I) diabetes mellitus: evidence from a study of extended haplotypes. Am. J. Hum. Diabetes Res. Clin. Pract. 24(Suppl), S95–S99 (1994).
Genet. 36, 1015–1023 (1984). 46. Mishra, R., Hodge, K. M., Cousminer, D. L., Leslie, R. D. & Grant, S. F. A. A global
12. Onengut-Gumuscu, S. et al. Type 1 diabetes risk in African-ancestry participants and perspective of latent autoimmune diabetes in adults. Trends Endocrinol. Metab. 29,
utility of an ancestry-specific genetic risk score. Diabetes Care 42, 406–415 (2019). 638–650 (2018).
13. Redondo, M. J., Steck, A. K. & Pugliese, A. Genetics of type 1 diabetes. Pediatr. Diabetes 47. Tosur, M. et al. Ethnic differences in progression of islet autoimmunity and type 1 diabetes
19, 346–353 (2018). in relatives at risk. Diabetologia 61, 2043–2053 (2018).
14. Chiou, J. et al. Interpreting type 1 diabetes risk with genetics and single-cell 48. Tosur, M. & Redondo, M. J. Heterogeneity of type 1 diabetes: the effect of ethnicity.
epigenomics. Nature 594, 398–402 (2021). Curr. Diabetes Rev. 14, 266–272 (2018).
15. Grant, S. F. A., Wells, A. D. & Rich, S. S. Next steps in the identification of gene targets 49. Redondo, M. J. et al. Racial/ethnic minority youth with recent-onset type 1 diabetes have
for type 1 diabetes. Diabetologia 63, 2260–2269 (2020). poor prognostic factors. Diabetes Care 41, 1017–1024 (2018).
16. Nekoua, M. P., Alidjinou, E. K. & Hober, D. Persistent coxsackievirus B infection and 50. Katte, J. C., McDonald, T. J., Sobngwi, E. & Jones, A. G. The phenotype of type 1 diabetes
pathogenesis of type 1 diabetes mellitus. Nat. Rev. Endocrinol. 18, 503–516 (2022). in sub-Saharan Africa. Front. Public Health 11, 1014626 (2023).
17. Girdhar, K. et al. A gut microbial peptide and molecular mimicry in the pathogenesis 51. Parviainen, A. et al. Heterogeneity of type 1 diabetes at diagnosis supports existence
of type 1 diabetes. Proc. Natl Acad. Sci. USA 119, e2120028119 (2022). of age-related endotypes. Diabetes Care 45, 871–879 (2022).
18. Blanter, M., Sork, H., Tuomela, S. & Flodstrom-Tullberg, M. Genetic and environmental 52. Leete, P. et al. The effect of age on the progression and severity of type 1 diabetes:
interaction in type 1 diabetes: a relationship between genetic risk alleles and molecular potential effects on disease mechanisms. Curr. Diabetes Rep. 18, 115 (2018).
traits of enterovirus infection? Curr. Diabetes Rep. 19, 82 (2019). 53. Casu, A. et al. Characteristics of adult- compared to childhood-onset type 1 diabetes.
19. Redondo, M. J. et al. A type 1 diabetes genetic risk score predicts progression of islet Diabet. Med. 37, 2109–2115 (2020).
autoimmunity and development of type 1 diabetes in individuals at risk. Diabetes Care 41, 54. Mishra, R. et al. Relative contribution of type 1 and type 2 diabetes loci to the genetic
1887–1894 (2018). etiology of adult-onset, non-insulin-requiring autoimmune diabetes. BMC Med. 15, 88
20. Ferrat, L. A. et al. A combined risk score enhances prediction of type 1 diabetes among (2017).
susceptible children. Nat. Med. 26, 1247–1255 (2020). 55. Cousminer, D. L. et al. First genome-wide association study of latent autoimmune
21. Ghalwash, M. et al. Two-age islet-autoantibody screening for childhood type 1 diabetes: diabetes in adults reveals novel insights linking immune and metabolic diabetes.
a prospective cohort study. Lancet Diabetes Endocrinol. 10, 589–596 (2022). Diabetes Care 41, 2396–2403 (2018).
22. Redondo, M. J. On the road to universal screening for risk of type 1 diabetes. 56. Howson, J. M. et al. Genetic analysis of adult-onset autoimmune diabetes. Diabetes 60,
Lancet Diabetes Endocrinol. 10, 554–555 (2022). 2645–2653 (2011).
23. Koskinen, M. K. et al. Longitudinal pattern of first-phase insulin response is associated 57. Thomas, N. J. et al. Frequency and phenotype of type 1 diabetes in the first six decades
with genetic variants outside the class II HLA region in children with multiple of life: a cross-sectional, genetically stratified survival analysis from UK Biobank.
autoantibodies. Diabetes 69, 12–19 (2020). Lancet Diabetes Endocrinol. 6, 122–129 (2018).
24. Evans-Molina, C. et al. β Cell dysfunction exists more than 5 years before type 1 diabetes 58. Redondo, M. J. et al. Single islet autoantibody at diagnosis of clinical type 1 diabetes
diagnosis. JCI Insight 3, e120877 (2018). is associated with older age and insulin resistance. J. Clin. Endocrinol. Metab. 105,
25. Shields, B. M. et al. C-peptide decline in type 1 diabetes has two phases: an initial 1629–1640 (2020).
exponential fall and a subsequent stable phase. Diabetes Care 41, 1486–1492 59. Leslie, R. D. et al. Adult-onset type 1 diabetes: current understanding and challenges.
(2018). Diabetes Care 44, 2449–2456 (2021).
26. Keenan, H. A. et al. Residual insulin production and pancreatic β-cell turnover after 60. So, M. et al. Advances in type 1 diabetes prediction using islet autoantibodies: beyond
50 years of diabetes: Joslin Medalist Study. Diabetes 59, 2846–2853 (2010). a simple count. Endocr. Rev. 42, 584–604 (2021).
27. Hagopian, W. et al. Teplizumab preserves C-peptide in recent-onset type 1 diabetes: 61. Ross, C. et al. The prevalence of islet autoantibodies in children and adolescents
two-year results from the randomized, placebo-controlled Protege trial. Diabetes 62, with type 1 diabetes mellitus: a global scoping review. Front. Endocrinol. 13, 815703
3901–3908 (2013). (2022).

Nature Reviews Endocrinology

Review article

62. Oram, R. A. et al. A type 1 diabetes genetic risk score can aid discrimination between 95. Redondo, M. J. et al. Milder loss of insulin-containing islets in individuals with type 1
type 1 and type 2 diabetes in young adults. Diabetes Care 39, 337–344 (2016). diabetes and type 2 diabetes-associated TCF7L2 genetic variants. Diabetologia 66,
63. Carlsson, A. et al. Absence of islet autoantibodies and modestly raised glucose values 127–131 (2023).
at diabetes diagnosis should lead to testing for MODY: lessons from a 5-year pediatric 96. Redondo, M. J., Grant, S. F., Davis, A., Greenbaum, C. & Biobank, T. D. E. Dissecting
Swedish National Cohort study. Diabetes Care 43, 82–89 (2020). heterogeneity in paediatric type 1 diabetes: association of TCF7L2 rs7903146 TT and
64. Balasubramanyam, A., Nalini, R., Hampe, C. S. & Maldonado, M. Syndromes of ketosis-prone low-risk human leukocyte antigen (HLA) genotypes. Diabet. Med. 34, 286–290 (2017).
diabetes mellitus. Endocr. Rev. 29, 292–302 (2008). 97. Cammidge, P. J. Diabetes mellitus and heredity. Br. Med. J. 2, 738–741 (1928).
65. Ilonen, J. et al. Associations between deduced first islet specific autoantibody with 98. Fajans, S. S. & Conn, J. W. Tolbutamide-induced improvement in carbohydrate tolerance
sex, age at diagnosis and genetic risk factors in young children with type 1 diabetes. of young people with mild diabetes mellitus. Diabetes 9, 83–88 (1960).
Pediatr. Diabetes 23, 693–702 (2022). 99. Tattersall, R. B. Mild familial diabetes with dominant inheritance. Q. J. Med. 43, 339–357
66. Krischer, J. P. et al. Characteristics of children diagnosed with type 1 diabetes before (1974).
vs after 6 years of age in the TEDDY cohort study. Diabetologia 64, 2247–2257 (2021). 100. Bell, G. I. et al. Gene for non-insulin-dependent diabetes mellitus (maturity-onset diabetes
67. Laine, A. P. et al. Non-HLA gene polymorphisms in the pathogenesis of type 1 diabetes: of the young subtype) is linked to DNA polymorphism on human chromosome 20q.
phase and endotype specific effects. Front. Immunol. 13, 909020 (2022). Proc. Natl Acad. Sci. USA 88, 1484–1488 (1991).
68. Arvan, P., Pietropaolo, M., Ostrov, D. & Rhodes, C. J. Islet autoantigens: structure, 101. Sperling, M. A. & Garg, A. in Diabetes in America Ch. 7 (eds Cowie, C. C. et al.) (National
function, localization, and regulation. Cold Spring Harb. Perspect. Med. 2, a007658 Institute of Diabetes and Digestive and Kidney Diseases, 2018).
(2012). 102. Kuruvilla, M. E., Lee, F. E. & Lee, G. B. Understanding asthma phenotypes, endotypes,
69. Giannopoulou, E. Z. et al. Islet autoantibody phenotypes and incidence in children and mechanisms of disease. Clin. Rev. Allergy Immunol. 56, 219–233 (2019).
at increased risk for type 1 diabetes. Diabetologia 58, 2317–2323 (2015). 103. Winthrop, K. L. et al. Unmet need in rheumatology: reports from the Advances
70. Hagan, D. W., Ferreira, S. M., Santos, G. J. & Phelps, E. A. The role of GABA in islet function. in Targeted Therapies meeting, 2022. Ann. Rheum. Dis. 82, 594–598 (2023).
Front. Endocrinol. 13, 972115 (2022). 104. Krogvold, L. et al. Pancreatic biopsy by minimal tail resection in live adult patients at the
71. Wyatt, R. C. et al. The first 142 amino acids of glutamate decarboxylase do not contribute onset of type 1 diabetes: experiences from the DiViD study. Diabetologia 57, 841–843
to epitopes recognized by autoantibodies associated with type 1 diabetes. Diabet. Med. (2014).
35, 954–963 (2018). 105. Morgan, N. G. & Richardson, S. J. Fifty years of pancreatic islet pathology in human type 1
72. Endesfelder, D. et al. Time-resolved autoantibody profiling facilitates stratification diabetes: insights gained and progress made. Diabetologia 61, 2499–2506 (2018).
of preclinical type 1 diabetes in children. Diabetes 68, 119–130 (2019). 106. Foulis, A. K., Liddle, C. N., Farquharson, M. A., Richmond, J. A. & Weir, R. S. The
73. Achenbach, P. et al. A classification and regression tree analysis identifies subgroups histopathology of the pancreas in type 1 (insulin-dependent) diabetes mellitus: a 25-year
of childhood type 1 diabetes. EBioMedicine 82, 104118 (2022). review of deaths in patients under 20 years of age in the United Kingdom. Diabetologia
74. Arif, S. et al. Blood and islet phenotypes indicate immunological heterogeneity in type 1 29, 267–274 (1986).
diabetes. Diabetes 63, 3835–3845 (2014). 107. Morgan, N. G., Richardson, S. J., Powers, A. C., Saunders, D. C. & Brissova, M. Images from
75. Arif, S. et al. Evaluating T cell responses prior to the onset of type 1 diabetes. Diabet. Med. the Exeter Archival Diabetes Biobank now accessible via Pancreatlas. Diabetes Care 45,
39, e14860 (2022). e174–e175 (2022).
76. Ferrara, C. T. et al. The role of age and excess body mass index in progression to type 1 108. Campbell-Thompson, M. et al. Network for pancreatic organ donors with diabetes
diabetes in at-risk adults. J. Clin. Endocrinol. Metab. 102, 4596–4603 (2017). (nPOD): developing a tissue biobank for type 1 diabetes. Diabetes Metab. Res. Rev. 28,
77. Redondo, M. J. et al. TCF7L2 genetic variants contribute to phenotypic heterogeneity 608–617 (2012).
of type 1 diabetes. Diabetes Care 41, 311–317 (2018). 109. Kaddis, J. S., Pugliese, A. & Atkinson, M. A. A run on the biobank: what have we learned
78. Ferrara, C. T. et al. Excess BMI in childhood: a modifiable risk factor for type 1 diabetes about type 1 diabetes from the nPOD tissue repository? Curr. Opin. Endocrinol. Diabetes
development? Diabetes Care 40, 698–701 (2017). Obes. 22, 290–295 (2015).
79. Censin, J. C. et al. Childhood adiposity and risk of type 1 diabetes: a Mendelian 110. Gepts, W. Pathologic anatomy of the pancreas in juvenile diabetes mellitus. Diabetes 14,
randomization study. PLoS Med. 14, e1002362 (2017). 619–633 (1965).
80. Redondo, M. J. et al. Index60 as an additional diagnostic criterion for type 1 diabetes. 111. In’t Veld, P. Insulitis in human type 1 diabetes: the quest for an elusive lesion. Islets 3,
Diabetologia 64, 836–844 (2021). 131–138 (2011).
81. Nathan, B. M. et al. Dysglycemia and Index60 as prediagnostic end points for type 1 112. Powers, A. C. Type 1 diabetes mellitus: much progress, many opportunities. J. Clin.
diabetes prevention trials. Diabetes Care 40, 1494–1499 (2017). Investig. https://doi.org/10.1172/JCI142242 (2021).
82. Sosenko, J. M. et al. Phenotypes associated with zones defined by area under the curve 113. Richardson, S. J. & Pugliese, A. 100 years of insulin: pancreas pathology in type 1
glucose and C-peptide in a population with islet autoantibodies. Diabetes Care 46, diabetes: an evolving story. J. Endocrinol. 252, R41–R57 (2021).
1098–1105 (2023). 114. Atkinson, M. A., Campbell-Thompson, M., Kusmartseva, I. & Kaestner, K. H. Organisation
83. Sims, E. K., Mirmira, R. G. & Evans-Molina, C. The role of beta-cell dysfunction in early of the human pancreas in health and in diabetes. Diabetologia 63, 1966–1973 (2020).
type 1 diabetes. Curr. Opin. Endocrinol. Diabetes Obes. 27, 215–224 (2020). 115. Leete, P. et al. Differential insulitic profiles determine the extent of β-cell destruction
84. Wu, H. & Ballantyne, C. M. Metabolic inflammation and insulin resistance in obesity. and the age at onset of type 1 diabetes. Diabetes 65, 1362–1369 (2016).
Circ. Res. 126, 1549–1564 (2020). 116. Inshaw, J. R. J., Cutler, A. J., Crouch, D. J. M., Wicker, L. S. & Todd, J. A. Genetic variants
85. Tao, L., Liu, H. & Gong, Y. Role and mechanism of the Th17/Treg cell balance in the predisposing most strongly to type 1 diabetes diagnosed under age 7 years lie near
development and progression of insulin resistance. Mol. Cell Biochem. 459, 183–188 candidate genes that function in the immune system and in pancreatic β-cells. Diabetes
(2019). Care 43, 169–177 (2020).
86. Butcher, M. J. et al. Association of proinflammatory cytokines and islet resident 117. Smith, M. J., Cambier, J. C. & Gottlieb, P. A. Endotypes in T1D: B lymphocytes and early
leucocytes with islet dysfunction in type 2 diabetes. Diabetologia 57, 491–501 (2014). onset. Curr. Opin. Endocrinol. Diabetes Obes. 27, 225–230 (2020).
87. Brooks-Worrell, B. M., Reichow, J. L., Goel, A., Ismail, H. & Palmer, J. P. Identification of 118. Leete, P. et al. Studies of insulin and proinsulin in pancreas and serum support the
autoantibody-negative autoimmune type 2 diabetic patients. Diabetes Care 34, 168–173 existence of aetiopathological endotypes of type 1 diabetes associated with age
(2011). at diagnosis. Diabetologia 63, 1258–1267 (2020).
88. Brooks-Worrell, B. et al. Islet autoimmunity is highly prevalent and associated with 119. Sims, E. K. et al. Proinsulin secretion is a persistent feature of type 1 diabetes.
diminished β-cell function in patients with type 2 diabetes in the grade study. Diabetes Diabetes Care 42, 258–264 (2019).
71, 1261–1271 (2022). 120. Rodriguez-Calvo, T. et al. Altered β-cell prohormone processing and secretion in type 1
89. Ferrara-Cook, C. et al. Excess BMI accelerates islet autoimmunity in older children diabetes. Diabetes 70, 1038–1050 (2021).
and adolescents. Diabetes Care 43, 580–587 (2020). 121. MacDonald, M. J. et al. A novel intron-encoded neuropilin-1 isoform in pancreatic islets
90. Krischer, J. P. et al. Predictors of the initiation of islet autoimmunity and progression associated with very young age of onset of type 1 diabetes. Diabetes 71, 2058–2063
to multiple autoantibodies and clinical diabetes: the TEDDY study. Diabetes Care 45, (2022).
2271–2281 (2022). 122. Ahlqvist, E. et al. Novel subgroups of adult-onset diabetes and their association with
91. Long, A. E. et al. Characteristics of slow progression to diabetes in multiple islet outcomes: a data-driven cluster analysis of six variables. Lancet Diabetes Endocrinol. 6,
autoantibody-positive individuals from five longitudinal cohorts: the SNAIL study. 361–369 (2018).
Diabetologia 61, 1484–1490 (2018). 123. Nishimura, A. et al. Slowly progressive type 1 diabetes mellitus: current knowledge
92. Krischer, J. P. et al. Genetic and environmental interactions modify the risk of diabetes- and future perspectives. Diabetes Metab. Syndr. Obes. 12, 2461–2477 (2019).
related autoimmunity by 6 years of age: the TEDDY study. Diabetes Care 40, 1194–1202 124. Libman, I. M. & Becker, D. J. Coexistence of type 1 and type 2 diabetes mellitus: “double”
(2017). diabetes? Pediatr. Diabetes 4, 110–113 (2003).
93. Redondo, M. J. et al. Transcription factor 7-like 2 (TCF7L2) gene polymorphism and 125. Brooks-Worrell, B. M. & Palmer, J. P. Attenuation of islet-specific T cell responses
progression from single to multiple autoantibody positivity in individuals at risk for type 1 is associated with C-peptide improvement in autoimmune type 2 diabetes patients.
diabetes. Diabetes Care 41, 2480–2486 (2018). Clin. Exp. Immunol. 171, 164–170 (2013).
94. Redondo, M. J. et al. Association of TCF7L2 variation with single islet autoantibody 126. Buzzetti, R., Pozzilli, P., Frederich, R., Iqbal, N. & Hirshberg, B. Saxagliptin improves
expression in children with type 1 diabetes. BMJ Open Diabetes Res. Care 2, e000008 glycaemic control and C-peptide secretion in latent autoimmune diabetes in adults
(2014). (LADA). Diabetes Metab. Res. Rev. 32, 289–296 (2016).

Nature Reviews Endocrinology

Review article

127. Agardh, C. D., Lynch, K. F., Palmer, M., Link, K. & Lernmark, A. GAD65 vaccination: 5 years 141. Mendes, D., Alves, C. & Batel-Marques, F. Number needed to treat (NNT) in clinical
of follow-up in a randomised dose-escalating study in adult-onset autoimmune diabetes. literature: an appraisal. BMC Med. 15, 112 (2017).
Diabetologia 52, 1363–1368 (2009). 142. Biomarkers Definitions Working Group. Biomarkers and surrogate endpoints:
128. Buzzetti, R. et al. Management of latent autoimmune diabetes in adults: a consensus preferred definitions and conceptual framework. Clin. Pharmacol. Ther. 69, 89–95
statement from an international expert panel. Diabetes 69, 2037–2047 (2020). (2001).
129. Fiorina, P. & Pozzilli, P. Unveiling a novel type 1 diabetes endotype: opportunities 143. Juneja, R. & Palmer, J. P. Type 1 1/2 diabetes: myth or reality? Autoimmunity 29, 65–83
for intervention. Diabetes Metab. Res. Rev. 38, e3536 (2022). (1999).
130. Jones, A. G., McDonald, T. J., Shields, B. M., Hagopian, W. & Hattersley, A. T. Latent
autoimmune diabetes of adults (LADA) is likely to represent a mixed population of Acknowledgements
autoimmune (type 1) and nonautoimmune (type 2) diabetes. Diabetes Care 44, 1243–1251 M.J.R.’s research is supported by the National Institutes of Health (NIH) grants R01 DK124395,
(2021). R01 DK121843 and U54 DK118638. Recent studies in N.G.M.’s laboratory have been supported
131. Redondo, M. J., Evans-Molina, C., Steck, A. K., Atkinson, M. A. & Sosenko, J. The influence by Diabetes UK, MRC, JDRF, EFSD and from the Innovative Medicines Initiative 2 Joint
of type 2 diabetes-associated factors on type 1 diabetes. Diabetes Care 42, 1357–1364 Undertaking under grant agreement no. 115797 (INNODIA) and no. 945268 (INNODIA
(2019). HARVEST). This Joint Undertaking receives support from the European Union’s Horizon 2020
132. Astudillo, M. et al. Type 2 diabetes in prepubertal children. Pediatr. Diabetes 22, 946–950 research and innovation programme, EFPIA, JDRF and The Leona M. and Harry B. Helmsley
(2021). Charitable Trust.
133. Libman, I. M. et al. Effect of metformin added to insulin on glycemic control among
overweight/obese adolescents with type 1 diabetes: a randomized clinical trial. Author contributions
J. Am. Med. Assoc. 314, 2241–2250 (2015). The authors contributed equally to all aspects of the article.
134. Cree-Green, M. et al. Metformin improves peripheral insulin sensitivity in youth with
type 1 diabetes. J. Clin. Endocrinol. Metab. 104, 3265–3278 (2019). Competing interests
135. Pescovitz, M. D. et al. B-lymphocyte depletion with rituximab and β-cell function: The authors declare no competing interests.
two-year results. Diabetes Care 37, 453–459 (2014).
136. Herold, K. C. et al. A single course of anti-CD3 monoclonal antibody hOKT3γ1(Ala-Ala) Additional information
results in improvement in C-peptide responses and clinical parameters for at least 2 years Peer review information Nature Reviews Endocrinology thanks Matthias von Herrath,
after onset of type 1 diabetes. Diabetes 54, 1763–1769 (2005). who co-reviewed with Estefania Masachs, Andrew Muir and Richard David Leslie for their
137. Hannelius, U., Beam, C. A. & Ludvigsson, J. Efficacy of GAD-alum immunotherapy contribution to the peer review of this work.
associated with HLA-DR3-DQ2 in recently diagnosed type 1 diabetes. Diabetologia 63,
2177–2181 (2020). Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims
138. Skyler, J. S. et al. Effects of oral insulin in relatives of patients with type 1 diabetes: in published maps and institutional affiliations.
the diabetes prevention trial – type 1. Diabetes Care 28, 1068–1076 (2005).
139. Writing Committee for the Type 1 Diabetes TrialNet Oral Insulin Study Group. Effect Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this
of oral insulin on prevention of diabetes in relatives of patients with type 1 diabetes: article under a publishing agreement with the author(s) or other rightsholder(s); author self-
a randomized clinical trial. J. Am. Med. Assoc. 318, 1891–1902 (2017). archiving of the accepted manuscript version of this article is solely governed by the terms
140. Sosenko, J. M. et al. Slowed metabolic decline after 1 year of oral insulin treatment of such publishing agreement and applicable law.
among individuals at high risk for type 1 diabetes in the diabetes prevention trial-type 1
(DPT-1) and TrialNet oral insulin prevention trials. Diabetes 69, 1827–1832 (2020). © Springer Nature Limited 2023

Nature Reviews Endocrinology