Misclassified latent autoimmune diabetes in adults within Māori and Pacific adults with type 2 diabetes in Aotearoa New Zealand

Correctly classifying diabetes is essential for ensuring appropriate and effective treatment strategies.¹ While type 1 diabetes mellitus (T1D) results from autoimmune destruction of pancreatic beta cells, leading to severe insulin deficiency, type 2 diabetes mellitus (T2D) is primarily characterised by insulin resistance.² This distinction has significant treatment implications, as T1D necessitates lifelong basal-bolus insulin therapy, while T2D is initially managed with lifestyle modifications and non-insulin medications, with later progression to simpler insulin regimens in some.³ However, the existence of latent autoimmune diabetes in adults (LADA), characterised by slowly progressive autoimmune beta cell destruction, adds complexity to this classification.²

Diagnosing LADA presents significant challenges due to its clinical similarity with features of both T1D and T2D.^4–6 The key definition of LADA includes adults diagnosed with diabetes at 30 years of age who test positive for T1D-associated autoantibodies and do not require insulin for at least 6 months after their diabetes diagnosis.^7,8 Those with a slowly progressive form of autoimmune diabetes at ages younger than 30 years are classified as latent autoimmune diabetes in the young (LADY).⁹ Those with LADY are likely to have more aggressive autoimmune destruction of beta cells and even more rapid progression to insulin requirement than LADA, given that youth with T2D have more rapid progression to insulin requirement than older adults with T2D.¹⁰ However, research comparing progression to insulin between LADY and LADA has not been done.

Studies, primarily conducted in European populations, show similarities in genetic risk scores (GRS) between LADA and T1D,^11,12 with the human leukocyte antigen (HLA) region playing a dominant role in T1D risk.¹³ This HLA influence on T1D GRS differentiation is also observed in Japanese and African ancestry populations, where incorporating both HLA and non-HLA variants enhances T1D risk prediction.^14,15 Interestingly, a European-derived T1D diagnostic model incorporating age, body mass index (BMI), autoantibodies (i.e., glutamic acid decarboxylase [GAD], islet antigen 2 [IA2] or zinc transporter 8 [ZnT8]) and a T1D GRS demonstrated good performance (area under the receiver operating characteristic curve [AUC] = 0.84) even in an Indian population, suggesting a broader applicability.^16,17

The prevalence of LADA has been reported to be high in Northern Europe and China compared to African American and Hispanic populations.^11,18,19 However, the prevalence among Māori and Pacific peoples living in Aotearoa New Zealand remains unknown. This study, leveraging data from the Genetics of Gout, Diabetes, and Kidney Disease (GoGDK) cohort,^20,21 aimed to: 1) determine the prevalence of LADA among Māori and Pacific adults clinically classified with T2D, 2) assess the distribution of a T1D GRS in those with T2D who did and did not have detectable autoantibodies, and 3) investigate any potential clinical differences, including glycaemic control and progression to insulin therapy, based on autoantibody presence or a high T1D GRS.

Methods

Study participants

All participants provided written informed consent for the collection of their samples and subsequent analyses. Ethical approval for this study was granted by the New Zealand Multi-Region Ethics Committee (MEC/05/10/130; MEC/10/09/092; MEC/11/04/036). The GoGDK study was initiated in 2005 to investigate the genetic and environmental contributors of gout, diabetes and kidney disease in adults from Aotearoa New Zealand. Participants were recruited through primary care clinics and community outreach programmes, aiming for broad population-level representation. Various clinical, demographic and biochemical measurements were collected from these participants at the time of recruitment. Longitudinal data for glucose lowering medications were collected from routine electronic healthcare dispensing records covering the period from 2002 to November 2022. A total of 2,538 participants who identified as Māori and/or Pacific from the GoGDK study of Aotearoa New Zealand were included in this analysis due to the availability of their genetic data for calculating their T1D GRS.^20,21 T2D diagnosis was ascertained in this cohort based on clinical records.

For this study, participants were reclassified with LADA according to the guidelines proposed by the Immunology of Diabetes Society (IDS): 1) over 30 years of age at diabetes onset, 2) positive titre for at least one T1D-associated autoantibody, such as GAD, IA2 and ZnT8, and 3) not treated with insulin the first 6 months after their diabetes diagnosis.²²

Autoantibody measurement

Autoantibodies associated with T1D, including GAD, IA2, and ZnT8, were measured simultaneously in available serum samples using the ElisaRSR^TM 3 Screen ICA^TM kit (RSR Limited, Cardiff, United Kingdom). Individuals were deemed positive for autoantibodies if their combined GAD/IA2/ZnT8 concentration was ≥20u/mL and/or if their sample yielded an index value ≥30. Samples were tested in duplicate, and a repeat test was conducted to confirm a positive autoantibody status.

Clinical diabetes characteristics

Demographic characteristics, BMI, diabetes, age of diabetes diagnosis, diabetes treatment and glycated haemoglobin (HbA_1c) were obtained from linked electronic health records retrieved from the TestSafe data repository. Access to this data was approved by the Regional Éclair Change Management Group. Individuals with diabetes were categorised into either a “high” or “low” clinical risk group for T1D based on their time to insulin initiation. Those who started insulin treatment within 3 years of their diabetes diagnosis were placed in the “high” risk group; all others were assigned to the “low” risk group. Achieving an HbA_1c ≤55mmol/mol indicated “good” control of blood glucose levels.

T1D GRS calculation

The T1D GRS consists of 30 variants from both HLA and non-HLA regions, accounting for DR3/DR4-DQ8 contribution.¹⁶ Specifically, there are five variants in the HLA region and 25 variants in other genes, including insulin (INS), interleukin genes (i.e., IL2, IL2RA, IL10, IL27), and protein kinase D2 (PRKD2). Assuming that non-DR3/DR4 risk alleles have a log-additive effect on T1D risk, the T1D GRS was calculated by summing the dosages across risk alleles multiplied by the weight (ln[odds ratio]) for each allele, divided by the number of variants (Appendix Table 1). HLA class II DR3/DR4-DQ8 haplotypes were inferred from two single-nucleotide polymorphisms (SNPs), rs2187668 and rs7454108, with corresponding weights assigned to each individual’s score.²² The whole-genome imputed sequence data constructed for GoGDK participants detailed in previous work was used to extract genotypes from a VCF file using BCFtools (v1.9-94-g9589876).^23,24 The ped files for these SNPs were imported into R v4.3.1 for processing and simple allelic scoring.²⁵ The Wellcome Trust Case Control Consortium GRS centiles were used as a guideline threshold, such that a T1D GRS >0.28 (>50th centile) suggests T1D.¹⁶

Statistical analysis

The Kruskal–Wallis test was employed to examine significant differences among group means of the T1D GRS. Subsequently, a Dunn’s test was used to identify pairwise groups with significant differences. A false discovery rate correction was applied for p-value adjustment. When comparing only the LADA and true T2D (autoantibody negative) groups, the Wilcoxon Rank-Sum Test was used for continuous variables, while the Pearson’s Chi-squared test was used for categorical variables. All analyses were performed using R software v4.3.1.²⁵

Results

Serum sample availability allowed testing for T1D autoantibodies in a subset of 293 participants: 262 with clinically classified T2D (age of diabetes diagnosis ranged 3–64 years) and 31 without T2D (age at testing 19–73 years). Among those with T2D, 23 (8.8%) tested positive for at least one autoantibody in the triple autoantibody panel (i.e., GAD/IA2/ZnT8). Of these, 14 (5.3% of the total T2D subset) met the criteria for LADA based on the IDS guidelines of age at diagnosis over 30 years, autoantibody positivity and absence of insulin treatment within the first 6 months of diagnosis (Table 1). Eight (3.1%) individuals that tested positive for autoantibodies met the criteria for LADY. Among the 31 controls without diabetes tested, one individual tested positive for autoantibodies and was recruited at 50 years of age, with a negative screening HbA_1c for diabetes at 58 years of age.

View Table 1–2, Figure 1–2.

Insulin initiation data were available for 10 of the 14 LADA participants. Of these, six (60%) eventually required insulin, with a mean time to insulin initiation of 13.75 (95% confidence interval [CI] 9.26–8.24) years. None of the 10 participants initiated insulin within the first 3 years of diagnosis, and the HbA_1c ranged from 34 to 96mmol/mol. Of the eight out of 10 who initiated insulin, these were at 9–26 years after their diabetes diagnosis. Seven of the eight participants with LADY initiated insulin within the observation period of 2002 to November 2022, with an average time to insulin of 18.29 years (95% CI 10.19–26.38). No significant differences in clinical diabetes characteristics (such as age at diagnosis, BMI, time to insulin, proportion on insulin or glycaemic control) were found between the LADA group and T2D group without T1D-associated autoantibodies.

The T1D GRS was calculated for a total of 2,538 participants, including 830 with clinically classified T2D and 1,708 without diabetes at the time of recruitment. Although 31 individuals met the criteria for “high” clinical risk for T1D due to insulin treatment within 3 years of diagnosis (age of diabetes diagnosis ranged from 12 to 78 years), none of these individuals tested positive for T1D autoantibodies. Among the 30 variants in the T1D GRS, 23 variants were present in the GoGDK cohort (Table 1). There was no significant difference in the T1D GRS between T1D autoantibody negative and positive groups (mean GRS [standard deviation] was 0.146 [0.016], n=269 and 0.148 [0.014], n=23, respectively; adjusted p=0.73; Figure 1) or between those with and without insulin treatment within 3 years of diagnosis (0.143 [0.017], n=31 and 0.147 [0.016], n=339, respectively; adjusted p=0.19; Figure 2 and Table 2).

Discussion

This study found that 5.3% (14/262) of the Māori and Pacific peoples in the GoGDK study that were clinically classified with T2D met the criteria for LADA. This was based on the presence of T1D-associated autoantibodies (GAD/IA2/ZnT8), age of diabetes diagnosis over 30 years and no insulin treatment within the first 6 months from diabetes diagnosis.²⁶ Additionally, 3.1% (8/262) of participants met the criteria for LADY, having tested positive for T1D-associated autoantibodies, age of diabetes diagnosis at an age younger than 30 years and not requiring insulin within the first 6 months of their diagnosis.¹¹ The LADA prevalence of 5.3% in our study is lower than the pooled global estimate of approximately 8.9%. Regional analysis shows that the Western Pacific Region has a prevalence of 8.3%, which is second lowest to Europe with a prevalence of 7%. This study is consistent with previous studies that have detected latent autoimmune diabetes in a subset of individuals diagnosed with T2D.^17,19,27,28 However, there were no significant differences in clinical characteristics, including the need for insulin therapy within 3 years of diagnosis, between those with LADA and those with autoantibody-negative T2D. Additional monitoring of C-peptide levels in the LADA and LADY group may have identified a sub-group with a rapid decline in beta cell function, who could have benefited from earlier insulin therapy.^8,29 While C-peptide measurements were not available for these participants, rapid decline in endogenous insulin secretion would be signalled by high HbA_1c, which is generally what is used in routine clinical practice decision making for when to initiate insulin therapy.

The observed prevalence of T1D-associated autoantibodies in the clinically ascertained T2D group was higher (8.8%; 23/262) than the false-positive rate observed in the control group (3.2%; 1/31). It has been suggested that identifying LADA in clinically defined T2D populations may lead to excess false-positive autoantibody results.⁵ Increasing the titre threshold for a single autoantibody and considering the presence of multiple autoantibodies may better align with clinical, biochemical and genetic characteristics of LADA/T1D.⁵ This requires further testing in larger cohorts of Māori and Pacific peoples with and without T1D or T2D, and soon after diagnosis since autoantibodies generally decline with time.³⁰

The T1D GRS aggregates the effects of multiple genetic variants associated with T1D risk and has shown promise in differentiating T1D from T2D in several populations,^16,31 but is less useful for differentiating LADA from T2D given the intermediate distribution of the T1D GRS in LADA between that seen in T1D and T2D.³² This study confirms the lack of utility of the T1D GRS in distinguishing LADA from T2D in this cohort of Māori and Pacific peoples. The absence of T1D cases in this study cohort means we cannot determine the utility of the T1D GRS assessment in Māori and Pacific peoples in distinguishing between T1D and T2D.

Our study has several limitations. The limited sample size, resulting in a small number of participants with LADA, may have limited our power to detect subtle differences in clinical characteristics or genetic risk. Serum samples were not available for all participants with T2D and genetic data to accurately test for the presence of T1D-associated autoantibodies. Additionally, autoantibody measurement on serum samples collected at recruitment, rather than at the time of diagnosis, could have underestimated the true prevalence of LADA due to the recognised decline of autoantibodies over time.³⁰ The use of the combined ElisaRSR^TM 3 Screen ICA^TM kit thresholds (combined GAD/IA2/ZnT8 concentration ≥20u/mL and/or ≥30 index value) is a potential limitation in not being able to differentiate titre or number of positive autoantibodies, given that lower titre single autoantibodies are more likely to be false positives than high titre of multiple autoantibodies. Finally, we recognise the absence of C-peptide measurements, which could better differentiate individuals with slowly progressive beta cell loss who may benefit from earlier insulin initiation, although in clinical practice HbA_1c is used to inform decisions around insulin initiation.³³

Despite these limitations, these findings contribute to the understanding of LADA in Māori and Pacific peoples. Larger studies with T1D GRS and comprehensive autoantibody testing at diagnosis of clinically classified T1D and T2D, and among controls, along with monitoring of C-peptide over time are warranted to clarify the utility of the T1D GRS and T1D autoantibody results in Māori and Pacific peoples. Such research will be crucial for effective strategies in diabetes prevention, screening and management for these populations.

View Appendix.