Monogenic forms of diabetes are estimated to account for around 1-4% of childhood diabetes, with distinct therapeutic implications from the more common type 1 and type 2 diabetes mellitus (1). However the diagnosis of monogenic diabetes remains a significant challenge and a study in the UK estimated >80% of monogenic diabetes is undiagnosed, largely due to variation in access for referrals for genetic testing (2).

Historically called Maturity Onset Diabetes of the Young (MODY), the classic triad was diagnosis under 25 years, a family history of diabetes with autosomal dominant inheritance and minimal/no requirement for insulin. However, an increasing list of monogenic diabetes genes and associated clinical features, combined with increased awareness and rising levels of type 2 diabetes, mean this triad is not as relevant today for selecting people who may benefit from genetic testing (3).

In Aotearoa New Zealand, there is publicly funded access to genetic testing for monogenic diabetes through endocrinologists. National guidelines for monogenic diabetes genetic testing from the New Zealand Society for the Study of Diabetes (NZSSD) require clinical identification of potentially affected individuals based on the absence of typical features for type 1 diabetes (i.e. absence of auto-antibodies, persistence of significant C-peptide) or type 2 diabetes (e.g. absence of obesity, hypertension, hyperlipidaemia) (4). In 2013, guidance was published on using a clinical phenotype-driven approach, where clinicians were directed to look for key features to identify single genes to test (5). For example, neonatal onset (directed neonatal diabetes gene panel testing); retrospective mild fasting hyperglycaemia (characteristic of GCK); high HDL (HNF1A); neonatal macrosomia (HNF4A) and syndromic features present in patients with either HNF1B or mitochondrial forms (MIDD) (5). In 2018, an updated guideline was released. Under this guideline, testing may be performed using single gene tests, if characteristic features are present for GCK, HNF1A/HNF4A, HNF1B, or mt.3243A>G; or alternatively, use a gene panel if characteristics are not clearly a single monogenic phenotype. This may be either as a small multi-gene panel (any combination of the 4 single genes) or a large gene panel, available overseas (6). Two different large gene panels are available – either a general monogenic diabetes panel, or a neonatal diabetes panel. The comparative diagnostic yield for monogenic diabetes testing using the clinical phenotype-guided single gene approach, compared to the small multi-gene panel testing approach or large gene panel testing has not previously been studied in the NZ population.

A clinical prediction model (MODY probability calculator) has been developed previously, using 594 Europeans with HNF1A, HNF4A or GCK monogenic diabetes and 597 with types 1 and type 2 diabetes, and can be used to derive the pre-test probability of monogenic diabetes (7). This calculator uses simple characteristics such as age at diagnosis, sex, time to insulin treatment, glucose lowering treatment, body mass index (BMI), HbA1c and family history of diabetes. The logistic regression model assumes a background prevalence of monogenic diabetes of 4.6% and can be used for individuals diagnosed with diabetes below 35 years of age and of European ethnicity. A >20% risk score for patients not on insulin, or >10% if insulin treated, is currently used in the United Kingdom, and referrals to the Exeter laboratory that include a calculation of the MODY probability score have a higher positive test rate than those that do not (33% vs 25%) (8). However, there is no data on the prevalence of monogenic diabetes in NZ and it is not known what MODY pre-test threshold would be optimal for use in the NZ population. This is particularly of note given the high prevalence of young-onset type 2 diabetes seen in people of Māori and Pacific ethnicities, making the NZ population clinically different to that of the UK (9).

Furthermore, the current testing algorithm relies on clinician judgement around atypical features for type 2 diabetes, which is more common and earlier in onset among Māori and Pacific peoples relative to people of European ethnicity. Hence, it is unclear whether testing volume and diagnostic rates for monogenic diabetes are different among Māori and Pacific people relative to other ethnicities.

Our first aim was to describe what is currently known about single gene versus small gene panel or large gene panel testing for monogenic diabetes in NZ, through evaluating the detection rate by gene and test type for all referrals for monogenic diabetes testing across NZ, between 2014 and 2021. Our second aim was to test whether the MODY probability calculator (7) could be used to guide patient selection for those without neonatal onset of diabetes or other syndromic forms of diabetes. Our third aim was to investigate whether testing frequency differed across ethnicities, particularly for Māori and Pacific populations. Fourthly, it is known that the highest yield for monogenic diabetes testing is in cascade testing, hence we aimed to examine the volume of cascade testing relative to proband testing. Finally, the information gathered in this audit was used to inform a new testing algorithm for monogenic diabetes in NZ.

All monogenic diabetes genetic test requests made by NZ endocrinologists between 01/01/2014 – 31/12/2020, were captured by searching the LabPlus laboratory (Auckland District Health Board, NZ) in house test database and the Exeter Genomics Laboratory (Royal Devon & Exeter Hospital, UK) in house database. Patients must have been suspected as having monogenic diabetes after assessment by a consultant endocrinologist and hence referred for genetic testing from any medical service location across NZ. Genetic testing for monogenic diabetes was publicly funded for all NZ residents.

Data for gene(s) tested, testing method, and result, as well as age, gender, and requestor was identified from test records for all patients. Where available, requestor provided information on BMI, age at diabetes diagnosis, history of diabetic ketoacidosis (DKA), time to insulin treatment, HbA1c and family history of a parent with diabetes was noted. Where sufficient clinical data was available, either on the request card or from the medical notes, and patients were aged 1-35 years at time of diagnosis, the MODY pre-test probability was calculated as per the model used by Shields et al. (7, 10).

LabPlus was the only laboratory in New Zealand offering testing of the genes GCK, HNF1A, HNF4A, or HNF1B during this period, allowing a comprehensive assessment of the state of monogenic diabetes testing within New Zealand. Test requests were performed either as a single gene test or small multi-gene panel test (any combination of the 4 genes listed above, performed in tandem).

In addition, requests sent to the Exeter laboratory were for either the neonatal diabetes gene panel or monogenic diabetes gene panel, both large panels as detailed below. Data on these referrals to the Exeter laboratory was available for patients sent from all regions across NZ. Requests for Monogenic Diabetes gene testing sent to laboratories other than Exeter were not captured but would be expected to be rare. Mitochondrial diabetes (MIDD) testing was not captured, as this was offered by another laboratory (Christchurch).

Referrals included a testing algorithm checklist (4, 5) where referrers indicated if single gene, small multigene, or large panel testing was requested, or cascade testing of a known gene variant. Whether patients had single gene testing, small multigene, or large panel testing, or any follow up testing was dependant on the clinician and hence this varied between patients with no standard pathway, other than the guidance detailed above.

Testing method at LabPlus was either full gene Sanger sequencing; cascade testing of a known familial variant (i.e. single exon Sanger sequencing only); or copy number (dosage) analysis using comparative genomic hybridisation (CGH) array.

For Sanger sequencing, the coding regions of the gene and flanking +/- 20bp (encompassing the splice sites) were sequenced. Analysis of sequence data was performed using Variant Reporter Software. Minimum sequence trace Phred score was 35, corresponding to an average false base call frequency of 0.031%.

Copy number analysis was performed using an Agilent 60K Custom CGH array for exonic deletions and duplications. Agilent CytoGenomics software was used for analysis. The inclusion threshold for analysis of a dosage change required a log2ratio of less than -0.25 over at least 3 contiguous probes for a deletion, and a log2ratio of greater than 0.25 over at least 3 contiguous probes for a duplication.

The Exeter Genomics Laboratory performed genetic testing for 27 monogenic diabetes genes that included the MT-TL1 m.3243A>G and 26 other autosomal diabetes genes (ABCC8, CEL, CISD2, GATA4, GATA6, GCK, HNF1A, HNF1B, HNF4A, INS, INSR, KCNJ11, LMNA, NEUROD1, PAX6, PCBD1, PDX1, PLIN1, POLD1, PPARG, RFX6, SLC29A3, TRMT10A, WFS1, ZBTB20, ZFP57). Patients under 12 months were tested for 30 neonatal diabetes genes (ABCC8, BSCL2, CISD2, CNOT1, COQ2, EIF2AK3, FOXP3, GATA4, GATA6, GCK, GLIS3, HNF1B, IER3IP1, IL2RA, INS, INSR, KCNJ11, LRBA, MNX1, NEUROD1, NEUROG3, NKX2-2, PDX1, PTF1A (coding and distal enhancer region), RFX6, SLC19A2, SLC2A2, STAT3, WFS1, ZFP57). The coding regions, 50 nucleotides of flanking intronic sequence of these genes and the mtDNA nucleotide m.3243 were analysed for single nucleotide variants (SNV), indels and gene deletions by targeted Next Generation Sequencing (tNGS). This assay did not target any other mitochondrial mutations or structural rearrangements. The Agilent SureSelect custom capture library and an Illumina NetSeq 500 NGS sequencing platform was used, according to the methodology described by Ellard et al. (6). The assay sequenced 99.7% of bases within the regions of interest at a minimum 30x read depth for all patients. Copy number variant (CNV) analysis was performed using ExomeDepth according to the methodology described by Parrish et al. (11). The estimated sensitivity for CNV detection was >95%. Neonatal diabetes cases also had Methylation-Specific Multiplex Ligation-Dependent Probe Amplication (MS-MLPA) for the 6q24 differentially methylated region (PLAGL1 DMR) using MRC Holland kit ME033.

Interpretation and classification of variants was based on the American College of Medical Genetics and Genomics (ACMG) guidelines (from July 2015 onwards) and on best available evidence prior to this. Results were classified by ACMG guidelines and reported as no variant detected (class 1-2 only), variant of uncertain significance (VUS, class 3, C3) or variant detected (likely pathogenic or pathogenic, class 4 or 5, C4/5 variant). Variants were described using the Human Genome Variation Society (HGVS) nomenclature guidelines. Results were classified as “positive” if a class 4 or 5 variant was identified, and “negative” or “normal” if a class 1, 2, or 3 variant were found.

Ethnicity data for all patients was coded using the Stats NZ classifications system, as European (Group 1), Māori (Group 2), Pacific Peoples (Group 3), Asian (Group 4), Middle Eastern/Latin American/African (MELAA) (Group 5) or other (Group 6 and 9). Respondents identifying as ‘Fijian Indian’ were coded as Asian; responders identifying as “New Zealander” were aggregated as European, as per the NZ ethnicity Data Protocols for the Health and Disability Sector. Prioritised output was used to assign those with multiple ethnic affiliations, with Māori and Pacific codes prioritised over European.

In total, 453 individual patients were tested at Auckland, of which 371 had proband/full gene sequencing (as opposed to cascade or dosage testing). A further 116 tests were performed at Exeter, of which 10 were the neonatal diabetes panel. Data was not available to determine which of the patients tested at Auckland went on to have further extended panel testing at Exeter, nor how many patients were sent straight to Exeter.

The detection rate for single gene testing and small multi-gene panels was relatively low, particularly compared with the detection rate of the NGS non-neonatal diabetes large gene panel used by Exeter (12% vs 23%). GCK sequencing, HNF1B dosage, and cascade testing offered the highest detection rates. Follow up testing after a negative single gene test was also mostly uninformative, with a detection rate of 4%, indicating the sequential single gene approach is inefficient. The referral rate for testing in patients identifying as Māori was similar to population data. However, Pacific people with diabetes were less likely to be selected for monogenic diabetes testing, compared to other ethnicities. Pacific people also had lower detection rates of monogenic diabetes (3% C4/C5 rate). The prevalence of diabetes by age shows the higher prevalence of obesity related type 2 diabetes in Pacific people, and hence the lower proportion of monogenic diabetes in this ethnic group are likely to be factors underpinning both reduced referral and diagnostic rates (13). However, it may represent other complex factors affecting patients’ presentations, interaction with healthcare providers, or clinician bias.

The prevalence of monogenic diabetes has not previously been reported in NZ. Compared to the Virtual Diabetes Register (VDR) data, (which lists 277,803 people with diabetes in NZ in 2020), the roughly 569 patients tested over this 7-year period represent just 0.2% of people with diabetes. If 1% prevalence is expected (a conservative estimate), this indicates a need to significantly increase testing rates. In other countries, work shows that the more testing is available, the higher the estimated prevalence of monogenic diabetes becomes (8).

Our findings are consistent with other countries’ data that GCK and HNF1A are the most common forms of monogenic diabetes. However, in the UK, HNF1A is more common (52%) than GCK (32%) (2) whilst in the many other countries, GCK is the most common (14–17). The reason for this difference is unknown – our finding that GCK is the most common form in NZ may either be a true difference to the UK population, or may indicate that the NZ testing strategy is resulting in easier access to GCK testing e.g. for pregnant women with gestational diabetes and children with hyperglycemia who are more often under specialist care, with insufficient testing of other patients with diabetes who are more often under primary care.

The high detection rate of HNF1B dosage testing (45%) likely reflects the more specific phenotype of HNF1B deletion patients, with syndromic and extra-diabetic features, and indicates single gene testing may still have a role in this setting. The MODY calculator is not validated for detection of HNF1B variants (7). Instead, the HNF1B score is recommended as a score of <8 can rule out HNF1B with 98% sensitivity, and a negative predictive value of >99% (18).

Dosage analysis of genes can generate significant extra workload for a laboratory, particularly in the single gene testing setting. Copy number variants (CNVs) have been documented to occur in the common MODY genes other than HNF1B but at low prevalence [1.2% HNF1A, 1.9% HNF4A, and 1.8% of GCK (19–21)]. In our study, in the small cohort of probands who received dosage analysis for genes other than HNF1B, no CNVs were found. This indicates that whilst dosage analysis is desirable, the added benefit to including this as routine for all patients is not high.

Data from 7 years of monogenic diabetes testing in NZ suggests more testing is to be encouraged, as testing rates are low, particularly for cascade genetic testing. A new testing strategy is proposed using an NGS based panel approach to improve diagnostic detection rates. A MODY probability score of >20% is proposed to be used to prioritise cost-effective selection of patients who are most likely to benefit from testing. Where a C4/C5 variant is identified, the patient and family should be referred for genetic counselling, to enable cascade testing of family members, an aspect which is currently significantly under-utilised.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the ADHB (Auckland District Health Board) research board (reference 7581). The study was assessed as being low risk and out of scope of HDEC (Health and Disability Ethics committees). Written informed consent from the participants’ legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.

RM and FH designed the study. FH analysed the data and prepared the manuscript. MG reviewed the manuscript and provided laboratory input. KC supplied data from Exeter. RP and RM reviewed the manuscript and provided clinical input. RM provided overall supervision. All authors contributed to the article and approved the submitted version.