Severe insulin-deficient diabetes: Updated review of epidemiology, biology, and clinical implications

Abstract

Type 2 diabetes mellitus is increasingly recognized as heterogeneous in pathophysiology, clinical course, and treatment needs. Data-driven clustering across diverse cohorts has identified severe insulin-deficient diabetes (SIDD), whose frequency varies by race/ethnicity and geography, with a higher relative burden reported in South Asian populations. SIDD is characterized by marked hyperglycemia, profound β-cell dysfunction, relatively lower or intermediate body mass index, and absence of autoimmune markers. Genetic and multi-omic evidence increasingly implicates insulin secretion pathways and β-cell stress programs rather than autoimmunity, supporting biological distinctiveness. Clinically, SIDD is consistently associated with adverse microvascular outcomes particularly diabetic retinopathy near diagnosis and during follow-up. Associations with macrovascular disease are more heterogeneous and may reflect differences in ascertainment, glycemic exposure, and treatment. Emerging evidence also suggests heterogeneity in all-cause mortality and excess life-years lost across diabetes and prediabetes subtypes, with insulin-deficient phenotypes showing an adverse mortality profile. Therapeutically, SIDD often requires early escalation and frequent insulin and may respond poorly to metformin alone. Key gaps include subtype stability over time and lack of cluster-guided trials; priorities include electronic health record-implementable classification and trials testing subtype-tailored strategies using hard outcomes.

Key Words: Type 2 diabetes; Severe insulin-deficient diabetes; Insulin deficiency; Insulin resistance; Beta-cell function

Core Tip: Severe insulin-deficient diabetes (SIDD) is a reproducible, non-autoimmune subtype of type 2 diabetes defined by profound β-cell dysfunction, marked hyperglycemia, and relatively lower body mass index. This review synthesizes global evidence showing that SIDD carries disproportionate microvascular risk particularly retinopathy and often requires early treatment intensification with frequent insulin use. A key innovation is highlighting population-specific subtype distributions, with insulin-deficient phenotypes more common in South Asians and linked to emerging signals for higher mortality and excess life-years lost. We integrate genetic and multi-omic insights implicating β-cell stress pathways and outline priorities for electronic health record-ready classification and subtype-guided trials.

INTRODUCTION

Type 2 diabetes mellitus (T2D) is heterogeneous in presentation, course, and response to therapy. A landmark clustering analysis in the Swedish ANDIS registry introduced five data-driven subtypes severe autoimmune diabetes, severe insulin-deficient diabetes (SIDD), severe insulin-resistant diabetes (SIRD), mild obesity-related diabetes (MOD), and mild age-related diabetes derived from six routinely collected variables [age at diagnosis, body mass index (BMI), glycosylated hemoglobin (HbA1c), homeostatic model assessment of beta-cell function (HOMA2-B), homeostatic model assessment of insulin resistance (HOMA2-IR), and glutamic acid decarboxylase (GAD) antibodies][1]. Among these, SIDD is distinguished by early onset, low/normal BMI, very high HbA1c and markedly reduced β-cell function in the absence of autoimmunity. Subsequent validation and perspective analyses have debated clinical utility but consistently recognize SIDD as a reproducible, clinically meaningful phenotype across cohorts[2-7]. This review synthesizes advances since the initial report, diversifies the evidentiary base across regions, and integrates recent insights.

LITERATURE IDENTIFICATION AND SELECTION

This review synthesizes evidence on SIDD published since the original clustering framework described by Ahlqvist et al[1]. Literature was identified through targeted searches of PubMed/MEDLINE, EMBASE, and Web of Science covering January 2018 through January 2026. Search terms included combinations of “severe insulin-deficient diabetes”, “SIDD”, “diabetes subtypes”, “data-driven diabetes clusters”, “Ahlqvist clusters”, “insulin-deficient diabetes”, and “precision diabetes”.

Studies were prioritized if they: (1) Used clustering or classification approaches derived from the original ANDIS framework; (2) Evaluated clinical characteristics, complications, or outcomes of SIDD or analogous insulin-deficient subtypes; or (3) Investigated genetic, molecular, or treatment implications of subtype classification. Both unsupervised clustering studies and studies applying supervised subtype assignment were included.

When multiple publications used overlapping cohorts, the most comprehensive or recent analyses were prioritized. Given the rapidly evolving literature and heterogeneity of study designs, this review is intended as a structured narrative synthesis rather than a formal systematic review, emphasizing studies with large sample sizes, longitudinal follow-up, or multi-population replication.

CLINICAL AND METABOLIC CHARACTERISTICS

Across studies, the definition of SIDD is not uniform. The original ANDIS framework uses unsupervised clustering based on six variables, including age at diagnosis, BMI, HbA1c, HOMA2-B, HOMA2-IR, and GAD antibodies[1]. Subsequent studies have applied supervised or nearest-centroid approaches using similar variables to assign individuals to predefined subtypes[6,7]. In settings where HOMA-derived indices, autoantibodies, or C-peptide measurements are unavailable, several studies have used simplified or “SIDD-like” proxy definitions based on routinely available clinical variables[8-12]. These differing approaches are not directly equivalent and should be considered when comparing subtype prevalence, biological interpretation, and outcome associations across cohorts.

SIDD’s cardinal feature is β-cell secretory failure without autoimmunity. In ANDIS, SIDD had the lowest HOMA2-B and highest HbA1c at diagnosis despite younger age and lower BMI than SIRD and MOD[1]. Independent cohorts have replicated these metabolic signatures: German new-onset diabetes[7], Asian Indian nationwide studies (INDIAB/INSPIRED and CARRS)[8,9], Mexican and Mexican-American populations[10,11], and United States NHANES analyses showing race/ethnicity-specific distributions[12] main characteristics (age, BMI and HbA1c) are summarized in Table 1. Chinese cohorts also identify SIDD (or analogous insulin-deficient clusters) soon after diagnosis; genetic and clinical correlates track with impaired β-cell traits rather than insulin resistance[3,5,13,14]. Across settings, SIDD is frequently negative for GAD and insulinoma-associated antigen-2 autoantibodies, differentiating it from latent autoimmune diabetes in adults despite superficially similar age and glycemia[4,15]. Collectively, these data support SIDD as a non-autoimmune, insulinopenic state within the T2D spectrum, likely reflecting intrinsic β-cell fragility.

Table 1 Summary of main characteristics of severe insulin-deficient diabetes across various races and ethnicities[1,5,7-10,12,14].

Ref.	Cohort	Setting	Age (years)	BMI (kg/m2)	HbA1c (%)
Ahlqvist et al[1]	ANDIS (Sweden)	Incident T2D	50.3	27.3	11.5
Zaharia et al[7]	German recent-onset	Recent-onset T2D (5-year follow-up)	43.8	27.0	8.7
Anjana et al[8]	INSPIRED/INDIAB (India)	< 5-year duration; multi-center	42.5	24.9	10.7
Jagannathan et al[9]	CARRS (India)	> 10-year duration, multi-center	49.8	27.1	10.4
Wang et al[14]	China	Newly diagnosed	49.6	25	9.15
Li et al[5]	China	Community T2D	58.09	25.44	9.53
Bello-Chavolla et al[10]	Mexican/Mexican-American	Mixed (NHANES + Mexican cohorts)	46.9	32.09	NA
Antonio-Villa et al[12]	United States NHANES	Population survey (classification)	58.17	31.18	9.9

NA: Not applicable; BMI: Body mass index; HbA1c: Glycosylated hemoglobin; T2D: Type 2 diabetes.

COMPLICATIONS, MORTALITY, AND LIFE-YEARS LOST

Interpretation of subtype-specific complication risks requires careful consideration of study design and cohort composition. In incident cohorts such as ANDIS and other recent-onset European studies, clustering variables largely reflect early disease physiology before substantial treatment exposure. In contrast, prevalent cohorts, including community-based or long-duration diabetes populations, may reflect metabolic adaptation, treatment intensification, or survival bias. Measures such as HbA1c, BMI, and fasting insulin or C-peptide may therefore partly capture treatment effects rather than underlying pathophysiology. These differences can influence subtype prevalence and the interpretation of complication risks. In addition, variation in cumulative glycemic exposure, treatment escalation, and screening intensity may confound or mediate observed associations between subtype classification and clinical outcomes.

Microvascular risk is consistently elevated in SIDD. In ANDIS, SIDD presented with the highest retinopathy burden near diagnosis; similar patterns were observed at 5-year follow-up in European cohorts[7] and in Asian Indian datasets[8]. Chinese community-based studies corroborate higher complication risks in insulin-deficient clusters vs milder phenotypes[5]. Macrovascular signals are more heterogeneous: Some reports suggest no excess vs SIRD after adjustment, while others indicate higher composite events when hyperglycemia is prolonged[7,14]. Importantly in South Asian populations, a recent analysis from the CARRS cohort reported heterogeneity in all-cause mortality and excess life-years lost across diabetes and prediabetes subtypes, with SIDD subtypes exhibiting the highest risk (hazard ratio = 3.34, 95% confidence interval: 2.39-4.68) and the greatest excess life-years lost[9].

TREATMENT TRAJECTORIES AND MANAGEMENT IMPLICATIONS

Therapeutically, SIDD trajectories are characterized by early pharmacologic escalation and frequent insulin requirement. In ANDIS and trial reanalyses, SIDD initiated glucose-lowering therapy and insulin earlier than other subtypes and was less likely to achieve durable control on metformin alone[1,6,7]. While the clusters were not superior to outcome-specific prediction models in some analyses, they provide a mechanistically coherent frame for stratified treatment hypotheses[7,16,17]. From a physiological perspective, SIDD may represent a state of relative insulin deficiency in which early strategies aimed at reducing β-cell stress and preserving residual function are particularly relevant. Incretin-based therapies and sodium glucose transporter 2 inhibitors may confer benefits through β-cell rest, weight neutrality, and cardiorenal protection, but cluster-stratified randomized data are lacking. Priority trials include early insulin vs glucagon-like peptide-1 receptor agonist-anchored strategies and the development of pragmatic electronic health record (EHR)-based algorithms that assign putative subtype from routine variables[17-19]. In real-world settings, simplified classifiers using variables such as age at diagnosis, BMI, HbA1c, fasting glucose, and, where available, fasting insulin or C-peptide may provide pragmatic approximations of insulin-deficient phenotypes. Surrogate markers such as the triglyceride-glucose index or high-density lipoprotein cholesterol have also been explored when HOMA-derived indices are unavailable, although such approaches require careful validation because reduced variable sets may increase misclassification and reduce subtype stability across populations.

MOLECULAR AND GENETIC UNDERPINNINGS

A 2025 meta-analysis across 19 studies (approximately 60000 cases) confirms that SIDD consistently exhibits lower β-cell function and higher HbA1c than other data-driven subgroups, strengthening external validity across ancestries[20]. Genetic dissection indicates that SIDD is enriched for alleles affecting insulin synthesis/secretion (e.g., MTNR1B, TCF7 L2, CDKAL1, HHEX) rather than autoimmune HLA loci[2,3,14,21]. Polygenic analyses partition diabetes into mechanism-weighted pathways, with SIDD aligning to β-cell deficiency signatures and distinct from insulin-resistant clusters. Multi-omics work (transcriptomic/epigenomic) highlights oxidative and endoplasmic reticulum stress programs in human islets and circulating tissues associated with insulin secretion impairment, though SIDD-specific datasets remain limited[21,22]. Emerging evidence also implicates gut microbial and host-microbial metabolic pathways in the pathophysiology of T2D, including effects on glycemic regulation, inflammation, and insulin sensitivity[23]. However, subtype-specific microbiome signatures in SIDD remain incompletely characterized and represent an important area for future investigation. These molecular signals suggest translational opportunities, including biomarkers of β-cell stress, pathway-informed therapeutic targets, and polygenic risk frameworks that may help identify individuals predisposed to insulin-deficient diabetes early in the disease course. They also provide a framework for investigating mechanisms of β-cell recovery and potential islet cell regeneration, particularly in insulin-deficient phenotypes such as SIDD, where profound β-cell dysfunction represents a central pathophysiological feature.

REPLICATION AND GENERALIZABILITY ACROSS POPULATIONS

Replications now span Europe (e.g., Denmark; Scotland/United Kingdom), North America, South Asia, East Asia, and the Gulf, and collectively demonstrate that the distribution of diabetes subtypes varies across populations. In the Scandinavian discovery cohort (ANDIS), SIDD accounted for 17.5% of adult-onset diabetes. In contrast, in South Asia, insulin-deficient phenotypes comprise a larger share: In the Indian INSPIRED analysis, 21% of participants were classified as SIDD, with an additional 12% in an insulin-deficient-predominant mixed subtype (CIRDD), consistent with greater β-cell deficiency in South Asian populations[8]. In the CARRS multi-cohort South Asian dataset (n = 2639 with T2D), SIDD represented 22.9%, while mild insulin-deficient diabetes represented 54.5% and SIRD 22.6%, further illustrating population-specific subtype composition in this region[9]. NHANES analyses also indicate variation in subtype frequency by race/ethnicity and sex in the United States, including relative enrichment of insulin-deficient phenotypes in some non-White groups[12]. In Mexico and Mexican-American samples, both supervised and unsupervised approaches identify an insulin-deficient cluster associated with early complications[10,11]. An Emirati etiological soft-clustering study in long-standing T2D similarly identified an insulin-deficient group with poor glycemic control, supporting external validity beyond incident-case cohorts while reinforcing geographic heterogeneity in subtype distributions[24]. Table 2 below summarizes the main clusters distribution and differences in SIDD age and BMI at diagnosis across three cohorts (ANDIS, INSPIRED, and CARRS).

Table 2 Distribution of data-driven diabetes subtypes across cohorts, with severe insulin-deficient diabetes age/body mass index at diagnosis, mean ± SD/n (%).

Ref.	Cohort	Population/setting	T2D analytic sample (n)	Subtype distribution	SIDD age at diagnosis	SIDD BMI at diagnosis
Ahlqvist et al[1]	ANDIS	Sweden (Europe)	8980	SIDD 1575 (17.5); SAID 577 (6.4); SIRD 1373 (15.3); MOD 1942 (21.6); MARD 3513 (39.1)	56.74 ± 11.14	28.86 ± 4.77
Anjana et al[8]	INSPIRED	India (South Asia)	19084	SIDD 5009 (26.2); IROD 4934 (25.9); CIRDD 2313 (12.1); MARD 6828 (35.8)	42.5 ± 10.8	24.9 ± 3.5
Jagannathan et al[9]	CARRS	South Asians (CARRS program)	2639	SIDD 605 (22.9); MIDD 1439 (54.5); SIRD 595 (22.5)	49.8	27.1 ± 3.5

BMI: Body mass index; T2D: Type 2 diabetes; SIDD: Severe insulin-deficient diabetes; SAID: Severe autoimmune diabetes; SIRD: Severe insulin-resistant diabetes; MOD: Mild obesity-related diabetes; MARD: Mild age-related diabetes; IROD: Insulin-resistant obese diabetes; CIRDD: Insulin-deficient-predominant mixed subtype; MIDD: Mild Insulin-deficient diabetes.

LIMITATIONS OF CURRENT EVIDENCE

Current limitations temper translational use. First, subtype stability remains an important challenge for clinical implementation. Longitudinal analyses suggest that a subset of individuals transition between clusters over time, particularly between SIDD and milder phenotypes as glycemic control improves or treatment intensifies. Potential drivers of reclassification include disease progression, progressive β-cell dysfunction, pharmacologic therapy, and weight change. These dynamics suggest that cluster assignment may represent a time-specific metabolic state rather than a fixed biological category and highlight the need for longitudinal validation and dynamic classification approaches[3,6]. Second, the reliance on HOMA-derived indices and GAD assays may complicate implementation in low-resource settings; simplified proxies (e.g., C-peptide, fasting insulin, high-density lipoprotein cholesterol, triglyceride-glucose index) warrant evaluation[18]. Third, absence of cluster-guided randomized trials leaves uncertainty about causal benefits of subtype-tailored therapy. Finally, harmonization across cohorts including variable definitions, timing (incident vs prevalent), and ancestry diversity remains essential for equitable precision diabetes medicine.

FUTURE DIRECTIONS

Key next steps center on three priorities. First, developing and openly releasing implementable subtype classifiers that can be computed directly from routinely captured EHR elements (e.g., age at diagnosis, BMI, HbA1c, fasting glucose, and where available C-peptide/insulin), with transparent feature definitions, transportable code, and prospective external validation across health systems. These tools should be coupled to routine recalibration and model-drift surveillance as laboratory platforms, prescribing patterns, and case-mix evolve[18]. Second embedding cluster-stratified randomization in first-line and second-line therapy trials to test early insulin vs incretin-anchored care in SIDD. Third, advancing integrated multi-omic and longitudinal cohort efforts to define biologically grounded phenotypes that extend beyond static clustering, with endpoints that include mortality, complications, patient-reported outcomes, and cost-effectiveness particularly in populations such as South Asians where insulin-deficient phenotypes are common[8,9,12].

CONCLUSION

SIDD is a reproducible, non-autoimmune insulinopenic subtype linked to severe hyperglycemia, early microvascular complications, and rapid treatment escalation across ancestries. While prognostic salience is clear, definitive evidence that SIDD-guided therapy improves outcomes is pending. With method harmonization and trials that operationalize subtypes at the point of care, SIDD offers a concrete pathway toward mechanisminformed precision diabetes especially salient in South Asian and other populations with high β-cell fragility[1,6-10,12,13,17,20,24].