Search

We aimed to compare the performance of approaches for classifying insulin treated diabetes within research datasets without measured classification biomarkers, evaluated against two independent biological definitions of diabetes type.We compared accuracy of ten reported approaches for classifying insulin treated diabetes into type 1 (T1D) and type 2 (T2D) diabetes in two cohorts: UK Biobank (UKBB) n=26,399 and DARE n=1,296. Overall performance for classifying T1D and T2D were assessed using: a T1D genetic risk score and genetic stratification method (UKBB); C-peptide measured at3 years diabetes duration (DARE).Approaches accuracy ranged from 71%-88% (UKBB) and 68%-88% (DARE). When classifying all participants, combining early insulin requirement with a T1D probability model (incorporating diagnosis age and BMI), and interview reported diabetes type (UKBB available in only 15%) consistently achieved high accuracy, (UKBB 87% and 87%, DARE 85% and 88% respectively). For identifying T1D with minimal misclassification, models with high thresholds or young diagnosis age (20 years) had highest performance. Findings were incorporated into an online tool identifying optimum approaches based on variable availability.Models combining continuous features with early insulin requirement are the most accurate methods for classifying insulin treated diabetes in research datasets without measured classification biomarkers.

Current treatment guidelines do not provide recommendations to support the selection of treatment for most people with type 2 diabetes. We aimed to develop and validate an algorithm to allow selection of optimal treatment based on glycaemic response, weight change, and tolerability outcomes when choosing between SGLT2 inhibitor or DPP-4 inhibitor therapies.In this retrospective cohort study, we identified patients initiating SGLT2 and DPP-4 inhibitor therapies after Jan 1, 2013, from the UK Clinical Practice Research Datalink (CPRD). We excluded those who received SGLT2 or DPP-4 inhibitors as first-line treatment or insulin at the same time, had estimated glomerular filtration rate (eGFR) of less than 45 mL/min per 1·73 mAmong 10 253 patients initiating SGLT2 inhibitors and 16 624 patients initiating DPP-4 inhibitors in CPRD, baseline HbAA validated treatment selection algorithm for SGLT2 inhibitor and DPP-4 inhibitor therapies can support decisions on optimal treatment for people with type 2 diabetes.BHF-Turing Cardiovascular Data Science Award and the UK Medical Research Council.

OBJECTIVE Recent joint American Diabetes Association and European Association for the Study of Diabetes guidelines recommend routine islet autoantibody testing in all adults newly diagnosed with type 1 diabetes. We aimed to assess the impact of routine islet autoantibody testing in this population. RESEARCH DESIGN AND METHODS We prospectively assessed the relationship between islet autoantibody status (GADA, IA-2A, and ZNT8A), clinical and genetic characteristics, and progression (annual change in urine C-peptide–to–creatinine ratio [UCPCR]) in 722 adults (≥18 years old at diagnosis) with clinically diagnosed type 1 diabetes and diabetes duration RESULTS Of 722 participants diagnosed with type 1 diabetes, 24.8% (179) were autoantibody negative. This group had genetic and C-peptide characteristics suggestive of a high prevalence of nonautoimmune diabetes: lower mean type 1 diabetes genetic risk score (islet autoantibody negative vs. positive: 10.85 vs. 13.09 [P < 0.001] [type 2 diabetes 10.12]) and lower annual change in C-peptide (UCPCR), −24% vs. −43% (P < 0.001). After median 24 months of follow-up, treatment change occurred in 36.6% (60 of 164) of autoantibody-negative participants: 22.6% (37 of 164) discontinued insulin, with HbA1c similar to that of participants continuing insulin (57.5 vs. 60.8 mmol/mol [7.4 vs. 7.7%], P = 0.4), and 14.0% (23 of 164) added adjuvant agents to insulin. CONCLUSIONS In adult-onset clinically diagnosed type 1 diabetes, negative islet autoantibodies should prompt careful consideration of other diabetes subtypes. When routinely measured, negative antibodies are associated with successful insulin cessation. These findings support recent recommendations for routine islet autoantibody assessment in adult-onset type 1 diabetes.

Aims/hypothesis Apparent type 2 diabetes is increasingly reported in lean adult individuals in sub-Saharan Africa. However, studies undertaking robust clinical and metabolic characterisation of lean individuals with new-onset type 2 diabetes are limited in this population. This cross-sectional study aimed to perform a detailed clinical and metabolic characterisation of newly diagnosed adult patients with diabetes in Uganda, in order to compare features between lean and non-lean individuals. Methods Socio-demographic, clinical, biophysical and metabolic (including oral glucose tolerance test) data were collected on 568 adult patients with newly diagnosed diabetes. Participants were screened for islet autoantibodies to exclude those with autoimmune diabetes. The remaining participants (with type 2 diabetes) were then classified as lean (BMI 2) or non-lean (BMI ≥25 kg/m2), and their socio-demographic, clinical, biophysical and metabolic characteristics were compared. Results Thirty-four participants (6.4%) were excluded from analyses because they were positive for pancreatic autoantibodies, and a further 34 participants because they had incomplete data. For the remaining 500 participants, the median (IQR) age, BMI and HbA1c were 48 years (39–58), 27.5 kg/m2 (23.6–31.4) and 90 mmol/mol (61–113) (10.3% [7.7–12.5]), respectively, with a female predominance (approximately 57%). Of the 500 participants, 160 (32%) and 340 (68%) were lean and non-lean, respectively. Compared with non-lean participants, lean participants were mainly male (60.6% vs 35.3%, ppppp Conclusions/interpretation Approximately one-third of participants with incident adult-onset non-autoimmune diabetes had BMI 2. Diabetes in these lean individuals was more common in men, and predominantly associated with reduced pancreatic secretory function rather than insulin resistance. The underlying pathological mechanisms are unclear, but this is likely to have important management implications. Graphical abstract

Patients with type 2 diabetes vary in their response to currently available therapeutic agents (including GLP-1 receptor agonists) leading to suboptimal glycemic control and increased risk of complications. We show that human carriers of hypomorphic T2D-risk alleles in the gene encoding peptidyl-glycine alpha-amidating monooxygenase(PAM),as well asPam-knockout mice, display increased resistance to GLP-1in vivo.Paminactivation in mice leads to reduced gastric GLP-1R expression and faster gastric emptying: this persists during GLP-1R agonist treatment and is rescued when GLP-1R activity is antagonized, indicating resistance to GLP-1’s gastric slowing properties. Meta-analysis of human data from studies examining GLP-1R agonist response (including RCTs) reveals a relative loss of 44% and 20% of glucose lowering (measured by glycated hemoglobin) in individuals with hypomorphicPAMalleles p.S539W and p.D536G treated with GLP-1R agonist. Genetic variation inPAMhas effects on incretin signaling that alters response to medication used commonly for treatment of T2D.(Funded by the Wellcome, Medical Research Council, European Union, NIHR Oxford Biomedical Research Centre, United Kingdom, Registered on ClinicalTrials.gov,NCT02723110.)Summary ParagraphType 2 diabetes (T2D) is a leading cause of morbidity and mortality globally1. Current management of T2D patients focuses on lowering glycemic exposure and reducing complications with lifestyle and pharmacological interventions2. Despite the availability of multiple medications to lower glycated hemoglobin (HbA1c), only 53% of individuals with T2D reach the glycemic target (HbA1c 3, 4. There is potential to improve medication selection through “precision medicine” where patient specific factors (e.g. genetic markers) are used to indicate whether a patient is more or less likely to respond to a medication. Here we show that human carriers of hypomorphic T2D-risk alleles in the gene encoding peptidyl-glycine alpha-amidating monooxygenase(PAM),as well asPam-knockout mice, have reduced PAM enzyme activity, display increased resistance to glucagon like peptide 1 (GLP-1)in vivoand have reduced response to the GLP-1 receptor agonist. Meta-analysis of human data from studies examining GLP-1 receptor agonist response (including RCTs) reveals a relative loss of 44% and 20% of glucose lowering (measured by glycated hemoglobin) in individuals with hypomorphicPAMalleles p.S539W and p.D536G treated with GLP-1 receptor agonist. Genetic variation inPAMhas effects on incretin signaling that alters response to medication used commonly for treatment of T2D.

AimsIn many low-income countries, fasting glucose is the primary measure of glycaemic control used for treatment titration, as HbA1c is often unavailable or unaffordable. Many patients in these countries walk long distances to the clinic, but the impact of walking on fasting glucose in type 2 diabetes is unknown. We aimed to determine whether this prolonged walking affects the reliability of fasting plasma glucose as a measure of glycaemic control.MethodsIn a randomised crossover trial, the change in glucose from baseline in the fasting state was compared between walking on a treadmill at a predetermined speed of 4.5 km/hour for 1 hour and not walking (resting) in people with type 2 diabetes. The pre-specified primary outcome was glucose at 1 and 2 hours.Results45 participants were enrolled and all completed both visits. 21/45 (46.7%) were female, and the median age was 51. Glucose during and after walking was similar to glucose while at rest; glucose difference (walking minus rest) was -0.15 (95% CI: -0.55, 0.26) and -0.10 (95% CI: - 0.50, 0.31) mmol/L at 1 and 2 hours respectively, p>0.4 for both).ConclusionsFasting plasma glucose is not meaningfully affected by prolonged walking in participants with type 2 diabetes; therefore, the reliability of fasting glucose for monitoring glycaemic burden is unlikely to be altered in patients who walk to the clinic.Research in contextWhat is already known about this subject?Fasting glucose is widely used to assess glycaemic control in people living with diabetes in low income countries, as HbA1c and home glucose monitoring are unaffordable. In these settings people living with diabetes will often walk long distances to receive healthcare.Little is known on the impact of walking on fasting glucose in people living with diabetes.What is the key question?Is fasting plasma glucose measure affected by a single bout of exercise such as walking in individuals with type 2 diabetes (T2D)?What are the new findings?There was no significant change in fasting glucose at the end of the walking exercise.There was no meaningful change in fasting glucose observed at any point up to 3 hours after commencing exercise.How might this impact on clinical practice in the foreseeable future?Fasting plasma glucose is not meaningfully affected by prolonged walking in participants with type 2 diabetes; therefore, the reliability of fasting glucose for monitoring glycaemic burden is unlikely to be altered in patients who walk to the clinic.

INTRODUCTION Type 1 diabetes in Africa has been associated with high mortality attributed mainly to poor insulin access. Free insulin provision programs for people with type 1 diabetes have been introduced across Africa recently. We aimed to determine the mortality rate and associated factors in a cohort of children and adolescents with type 1 diabetes who receive free insulin treatment in sub-Saharan Africa. METHODS We conducted a retrospective analysis using the Changing Diabetes in Children (CDiC) medical records in Cameroon between 2011 and 2015. RESULTS The overall mortality rate was 33.0 per 1000 person-years (95% CI 25.2-43.2). Most deaths (71.7%) occurred outside of the hospital setting, and the cause of death was known only in 13/53 (24.5%). Mortality was substantially higher in CDiC participants followed up in regional clinics compared to the main urban CDiC clinic in Yaounde; 41 per 1000 years (95% CI 30.8-56.0) versus 17.5 per 1000 years (95% CI 9.4-32.5), and in those with no formal education compared to those who had some level of education; 68.0 per 1000 years (95% CI 45.1-102.2) versus 23.6 per 1000 years (95% CI 16.5-33.8). In Cox proportional multivariable analysis, urban place of care (HR = 0.23, 95% CI 0.09-0.57; p = 0.002) and formal education (HR = 0.42, 95% CI 0.22-0.79; p = 0.007) were independently associated with mortality. CONCLUSION Despite free insulin provision, mortality remains high in children and adolescents with type 1 diabetes in Cameroon and is substantially higher in rural settings and those with no formal education.

The phenotype of type 1 diabetes in Africa, especially sub-Saharan Africa, is poorly understood. Most previously conducted studies have suggested that type 1 diabetes may have a different phenotype from the classical form of the disease described in western literature. Making an accurate diagnosis of type 1 diabetes in Africa is challenging, given the predominance of atypical diabetes forms and limited resources. The peak age of onset of type 1 diabetes in sub-Saharan Africa seems to occur after 18–20 years. Multiple studies have reported lower rates of islet autoantibodies ranging from 20 to 60% amongst people with type 1 diabetes in African populations, lower than that reported in other populations. Some studies have reported much higher levels of retained endogenous insulin secretion than in type 1 diabetes elsewhere, with lower rates of type 1 diabetes genetic susceptibility and HLA haplotypes. The HLA DR3 appears to be the most predominant HLA haplotype amongst people with type 1 diabetes in sub-Saharan Africa than the HLA DR4 haplotype. Some type 1 diabetes studies in sub-Saharan Africa have been limited by small sample sizes and diverse methods employed. Robust studies close to diabetes onset are sparse. Large prospective studies with well-standardized methodologies in people at or close to diabetes diagnosis in different population groups will be paramount to provide further insight into the phenotype of type 1 diabetes in sub-Saharan Africa.

Background: In the treatment of type 2 diabetes, GLP-1 receptor agonists lower blood glucose concentrations, body weight, and have cardiovascular benefits. The efficacy and side effects of GLP-1 receptor agonists vary between people. Human pharmacogenomic studies of this inter-individual variation can provide both biological insight into drug action and provide biomarkers to inform clinical decision making. We therefore aimed to identify genetic variants associated with glycaemic response to GLP-1 receptor agonist treatment. Methods: In this genome-wide analysis we included adults (aged ≥18 years) with type 2 diabetes treated with GLP-1 receptor agonists with baseline HbA 1c of 7% or more (53 mmol/mol) from four prospective observational cohorts (DIRECT, PRIBA, PROMASTER, and GoDARTS) and two randomised clinical trials (HARMONY phase 3 and AWARD). The primary endpoint was HbA 1c reduction at 6 months after starting GLP-1 receptor agonists. We evaluated variants in GLP1R, then did a genome-wide association study and gene-based burden tests. Findings: 4571 adults were included in our analysis, of these, 3339 (73%) were White European, 449 (10%) Hispanic, 312 (7%) American Indian or Alaskan Native, and 471 (10%) were other, and around 2140 (47%) of the participants were women. Variation in HbA 1c reduction with GLP-1 receptor agonists treatment was associated with rs6923761G→A (Gly168Ser) in the GLP1R (0·08% [95% CI 0·04–0·12] or 0·9 mmol/mol lower reduction in HbA 1c per serine, p=6·0 × 10 −5) and low frequency variants in ARRB1 (optimal sequence kernel association test p=6·7 × 10 −8), largely driven by rs140226575G→A (Thr370Met; 0·25% [SE 0·06] or 2·7 mmol/mol [SE 0·7] greater HbA 1c reduction per methionine, p=5·2 × 10 −6). A similar effect size for the ARRB1 Thr370Met was seen in Hispanic and American Indian or Alaska Native populations who have a higher frequency of this variant (6–11%) than in White European populations. Combining these two genes identified 4% of the population who had a 30% greater reduction in HbA 1c than the 9% of the population with the worse response. Interpretation: This genome-wide pharmacogenomic study of GLP-1 receptor agonists provides novel biological and clinical insights. Clinically, when genotype is routinely available at the point of prescribing, individuals with ARRB1 variants might benefit from earlier initiation of GLP-1 receptor agonists.

ObjectiveTo compare individualized treatment selection strategies based on predicted individual-level treatment effects from a causal forest machine learning algorithm and a penalized regression model.Study Design and SettingCohort study characterizing individual-level glucose-lowering response (6 month reduction in HbA1c) in people with type 2 diabetes initiating SGLT2-inhibitor or DPP4-inhibitor therapy. Model development set comprised 1,428 participants in the CANTATA-D and CANTATA-D2 trials (SGLT2-inhibitor versus DPP4-inhibitor). For external validation, calibration of observed versus predicted differences in HbA1c in patient strata defined by size of predicted HbA1c benefit was evaluated in 18,741 UK primary care patients (Clinical Practice Research Datalink).ResultsHeterogeneity in treatment effects was detected in trial participants with both approaches (causal forest: 98.6% & penalized regression: 81.7% predicted to have a benefit on SGLT2-inhibitor therapy over DPP4-inhibitor therapy). In validation, calibration was good with penalized regression but sub-optimal with causal forest. A strata with an HbA1c benefit >10 mmol/mol with SGLT2-inhibitors (3.7% of patients, observed benefit 11.0 mmol/mol [95%CI 8.0-14.0]) was identified using penalized regression but not causal forest, and a much larger strata with an HbA1c benefit 5-10 mmol with SGLT2-inhibitors was identified with penalized regression (regression: 20.9% of patients, observed benefit 7.8 mmol/mol (95%CI 6.7-8.9); causal forest 11.6%, observed benefit 8.7 mmol/mol (95%CI 7.4-10.1).ConclusionWhen evaluating treatment effect heterogeneity researchers should not rely on causal forest (or other similar machine learning algorithms) alone, and must compare outputs with standard regression.What is new?QuestionWhat is the comparative utility of machine learning compared to standard regression for identifying variation in patient-level outcomes (treatment effect heterogeneity) due to different treatments?FindingsCausal forest and penalized regression models were developed using trial data to predict glycated hemoglobin [HbA1c]) outcomes with SGLT2-inhibitor and DPP4-inhibitor therapy in 1,428 individuals with type 2 diabetes. In external validation (18,741 patients), penalized regression outperformed causal forest at identifying population strata with a superior glycemic response to SGLT2-inhibitors compared to DPP4-inhibitors.ImplicationsStudies estimating treatment effect heterogeneity should not solely rely on machine learning and should compare results with standard regression.

Objective Recent joint American and European diabetes association guidelines recommend routine islet autoantibody testing in all adults newly diagnosed with type 1 diabetes. We aimed to assess the impact of routine islet autoantibody testing in this population. Research Design and Methods We prospectively assessed the characteristics and progression (annual change in Urine C-peptide Creatinine Ratio (UCPCR)) associated with islet autoantibody status (GAD, IA-2 and ZNT8) in 722 adults (≥ 18 years old at diagnosis) with clinically diagnosed type 1 diabetes and duration Results 24.8% (179/722) of participants diagnosed with type 1 diabetes were autoantibody negative. This group had genetic and C-peptide characteristics suggestive of a high prevalence of non-autoimmune diabetes: lower mean type 1 diabetes genetic risk score (islet autoantibody negative versus positive: 10.85 vs 13.09 (pp After median 24-months follow up, treatment change occurred in 36.6% (60/164) of autoantibody negative participants: 22.6% (37/164) discontinued insulin, with a HbA1c similar to those continuing insulin (57.5 vs 60.8mmol/mol [7.4 vs 7.7%], p=0.4) and 14.0% (23/164) added adjuvant agents to insulin. Conclusions In adult-onset clinically diagnosed type 1 diabetes, negative islet autoantibodies should prompt careful consideration of other diabetes subtypes. When routinely measured negative antibodies are associated with successful insulin cessation. These findings support recent recommendations for routine islet autoantibody assessment in adult-onset type 1 diabetes.

Precision medicine aims to treat an individual based on their clinical characteristics. A differential drug response, critical to using these features for therapy selection, has never been examined directly in type 2 diabetes. In this study, we tested two hypotheses: (1) individuals with body mass index (BMI) 30 kg/m

Positivity for islet autoantibodies is used for diagnosis of type 1 diabetes. However, the importance of the autoantibody level at diagnosis of type 1 diabetes is not clear. Here, we assessed the association of glutamate decarboxylase (GADA), islet antigen-2 (IA-2A) and zinc transporter 8 (ZnT8A) autoantibody levels, measured using radiobinding assays, on genetic and clinical characteristics at diagnosis of 1536 participants with diabetes who were positive for these autoantibodies. We show that GADA and IA-2A levels had bimodal distributions, but ZnT8A level did not. The comparison of genetic and clinical characteristics between high and low level categories showed high GADA level was associated with older age at diagnosis, female sex and HLA-DR3-DQ2, whereas high IA-2A level was associated with younger age of diagnosis, ZnT8A positivity and HLA-DR4-DQ8. We replicated our findings in an independent cohort of 427 people with type 1 diabetes where autoantibodies were measured using enzyme-linked immunosorbent assays. In conclusion, Islet autoantibody levels provide additional information over positivity in type 1 diabetes at diagnosis. The bimodality of islet autoantibody levels highlights the novel aspect of heterogeneity of type 1 diabetes which may have implications on prediction, treatment and prognosis.Islet autoantibodies are commonly used in the diagnosis and prediction of type 1 diabetes. They are well established as the biomarkers of the underlying autoimmune pathogenesis (1). Autoantibodies to islet cell antigen (ICA), glutamate decarboxylase (GADA), islet antigen-2 (IA-2A), insulin (IAA) and zinc transporter 8 (ZnT8A) are the most commonly used islet autoantibodies at diagnosis (2). As detectable islet autoantibodies overlap between health and disease, a test is usually considered positive for a given islet autoantibody when the antibody level is higher than a 97.5–99th centile of a control population (3; 4). In routine clinical practice, quantitative islet autoantibody results are usually interpreted as positive or negative, and the level of the islet autoantibody, is not thought to be clinically meaningful.Islet autoantibody levels may provide additional information over positivity in type 1 diabetes at diagnosis. Similar to type 1 diabetes, autoantibodies to a specific antigen are commonly used for diagnosis in many other autoimmune diseases (such as TSH receptor antibodies in Graves’ disease and Rheumatoid Factor and Citrullinated Protein in rheumatoid arthritis). For Graves’ disease and rheumatoid arthritis, along with autoantibody positivity for these antigens, autoantibody level at diagnosis is associated with disease severity, prognosis and treatment success (5; 6). Multiple studies have shown a role for islet autoantibody level in the prediction of onset of type 1 diabetes, those with a higher levels of IA-2A, IAA and ICA have an increased risk of developing type 1 diabetes in at-risk populations (1; 7-9). However, it is not clear if the islet autoantibody level at diagnosis of type 1 diabetes, in addition to its interpretation as ‘positive’, is associated with the clinical phenotype similar to other autoimmune diseases.In this study, we undertook an analysis of GADA, IA-2A and ZnT8A levels at diagnosis in a large cohort of participants with type 1 diabetes, assessing the association of islet autoantibody levels on genetic and clinical characteristics at diagnosis in people with type 1 diabetes.

AimsPopulation datasets are increasingly used to study type 1 or 2 diabetes, and inform clinical practice. However, correctly classifying diabetes type, when insulin treated, in population datasets is challenging. Many different approaches have been proposed, ranging from simple age or BMI cut offs, to complex algorithms, and the optimal approach is unclear. We aimed to compare the performance of approaches for classifying insulin treated diabetes for research studies, evaluated against two independent biological definitions of diabetes type.MethodWe compared accuracy of thirteen reported approaches for classifying insulin treated diabetes into type 1 and type 2 diabetes in two population cohorts with diabetes: UK Biobank (UKBB) n=26,399 and DARE n=1,296. Overall accuracy and predictive values for classifying type 1 and 2 diabetes were assessed using: 1) a type 1 diabetes genetic risk score and genetic stratification method (UKBB); 2) C-peptide measured at >3 years diabetes duration (DARE).ResultsAccuracy of approaches ranged from 71%-88% in UKBB and 68%-88% in DARE. All approaches were improved by combining with requirement for early insulin treatment (ConclusionSelf-reported diagnosis and models combining continuous features with early insulin requirement are the most accurate methods of classifying insulin treated diabetes in research datasets without measured classification biomarkers.

ObjectiveTo establish whether clinical patient characteristics routinely measured in primary care can identify people with differing short-term benefits and risks for SGLT2-inhibitor and DPP4-inhibitor therapies, and to derive and validate a treatment selection algorithm to identify the likely optimal therapy for individual patients.DesignProspective cohort study.SettingRoutine clinical data from United Kingdom general practice (Clinical Practice Research Datalink [CPRD]), and individual-level clinical trial data from 14 multi-country trials of SGLT2-inhibitor and DPP4-inhibitor therapies.Participants26,877 new users of SGLT2-inhibitor and DPP4-inhibitor therapy in CPRD over 2013-2019, and 10,414 participants randomised to SGLT2-inhibitor or DPP4-inhibitor therapy in 14 clinical trials, including 3 head-to-head trials of the two therapies (n=2,499).Main outcome measuresThe primary outcome was achieved HbA1c 6 months after initiating therapy. Clinical features associated with differential HbA1c outcomes with SGLT2-inhibitor and DPP4-inhibitor therapies were identified in routine clinical data, with associations then tested in trial data. A multivariable treatment selection algorithm to predict differential HbA1c outcomes was developed in a CPRD derivation cohort (n=14,069), with validation in a CPRD validation cohort (n=9,376) and the head-to-head trials. In CPRD, we further explored the relationship between model predictions and secondary outcomes of weight loss and treatment discontinuation.ResultsThe final treatment selection algorithm included HbA1c, eGFR, ALT, age, and BMI, which were identified as predictors of differential HbA1c outcomes with SGLT2-inhibitor and DPP4-inhibitor therapies using both routine and trial data. In validation cohorts, patient strata predicted to have a ≥5 mmol/mol HbA1c reduction with SGLT2-inhibitor therapy compared with DPP4-inhibitor therapy (38.8% of CPRD validation sample) had an observed greater reduction of 8.8 mmol/mol [95%CI 7.8-9.8] in the CPRD validation sample, a 5.8 mmol/mol (95%CI 3.9-7.7) greater reduction in the Cantata D/D2 trials, and a 6.6 mmol/mol [95%CI 2.2-11.0]) greater reduction in the BI1245.20 trial. In CPRD, there was a greater weight reduction with SGLT2-inhibitor therapy regardless of predicted glycaemic benefit. Strata predicted to have greater reduction in HbA1c on SGLT2-inhibitor therapy had a similar risk of discontinuation as on DPP4-inhibitor therapy. In contrast, strata predicted to have greater reduction in HbA1c with DPP4-inhibitor therapy were half as likely to discontinue DPP4-inhibitor therapy than SGLT2-inhibitor therapy.ConclusionsRoutinely measured clinical features are robustly associated with differential glycaemic responses to SGLT2-inhibitor and DPP4-inhibitor therapies. Combining features into a treatment selection algorithm can inform clinical decisions concerning optimal type 2 diabetes treatment choices.Key messagesWhat is already known on this subjectDespite there being multiple glucose-lowering treatment options available for people with type 2 diabetes, current guidelines do not provide clear advice on selecting the optimal treatment for most patients.It is unknown whether routinely measured clinical features modify the risks and benefits of two common treatment options, DPP4-inhibitor or SGLT2-inhibitor therapy, and which could be used to target these treatments to those patients most likely to benefit.What this study addsUsing data from 10,414 participants in 14 randomised trials, and 26,877 patients in UK primary care, we show several routinely available clinical features, notably glycated haemoglobin (HbA1c) and kidney function, are robustly associated with differential HbA1c responses to initiating SGLT2-inhibitor and DPP4-inhibitor therapies.Combining clinical features into a multivariable treatment selection model identifies validated patient strata with 1) a >5 mmol/mol HbA1c benefit for SGLT2-i therapy compared with DPP4-inhibitor therapy ; 2) a 50% reduced risk of early treatment discontinuation with DPP4-inhibitor therapy compared with SGLT2-inhibitor therapy.Our findings demonstrate a precision medicine approach based on routine clinical features can inform clinical decisions concerning optimal type 2 diabetes treatment choices.

IntroductionThe utility of HbA1c (glycosylated hemoglobin) to estimate glycemic control in populations of African and other low-resource countries has been questioned because of high prevalence of other medical conditions that may affect its reliability. Using continuous glucose monitoring (CGM), we aimed to determine the comparative performance of HbA1c, fasting plasma glucose (FPG) (within 5 hours of a meal) and random non-fasting glucose (RPG) in assessing glycemic burden.Research design and methodsWe assessed the performance of HbA1c, FPG and RPG in comparison to CGM mean glucose in 192 Ugandan participants with type 2 diabetes. Analysis was undertaken in all participants, and in subgroups with and without medical conditions reported to affect HbA1c reliability. We then assessed the performance of FPG and RPG, and optimal thresholds, in comparison to HbA1c in participants without medical conditions thought to alter HbA1c reliability.Results32.8% (63/192) of participants had medical conditions that may affect HbA1c reliability: anemia 9.4% (18/192), sickle cell trait and/or hemoglobin C (HbC) 22.4% (43/192), or renal impairment 6.3% (12/192). Despite high prevalence of medical conditions thought to affect HbA1c reliability, HbA1c had the strongest correlation with CGM measured glucose in day-to-day living (0.88, 95% CI 0.84 to 0.91), followed by FPG (0.82, 95% CI 0.76 to 0.86) and RPG (0.76, 95% CI 0.69 to 0.81). Among participants without conditions thought to affect HbA1c reliability, FPG and RPG had a similar diagnostic performance in identifying poor glycemic control defined by a range of HbA1c thresholds. FPG of ≥7.1 mmol/L and RPG of ≥10.5 mmol/L correctly identified 78.2% and 78.8%, respectively, of patients with an HbA1c of ≥7.0%.ConclusionsHbA1c is the optimal test for monitoring glucose control even in low-income and middle-income countries where medical conditions that may alter its reliability are prevalent; FPG and RPG are valuable alternatives where HbA1c is not available.

Type 2 diabetes (T2D) and non-alcoholic fatty liver disease (NAFLD) often co-occur. Defining causal pathways underlying this relationship may help optimize the prevention and treatment of both diseases. Thus, we assessed the strength and magnitude of the putative causal pathways linking dysglycemia and fatty liver, using a combination of causal inference methods.Measures of glycemia, insulin dynamics, magnetic resonance imaging (MRI)-derived abdominal and liver fat content, serological biomarkers, lifestyle, and anthropometry were obtained in participants from the IMI DIRECT cohorts (n=795 with new onset T2D and 2234 individuals free from diabetes). UK Biobank (n=3641) was used for modelling and replication purposes. Bayesian networks were employed to infer causal pathways, with causal validation using two-sample Mendelian randomization.Bayesian networks fitted to IMI DIRECT data identified higher basal insulin secretion rate (BasalISR) and MRI-derived excess visceral fat (VAT) accumulation as the features of dysmetabolism most likely to cause liver fat accumulation; the unconditional probability of fatty liver (>5%) increased significantly when conditioning on high levels of BasalISR and VAT (by 23%, 32% respectively; 40% for both). Analyses in UK Biobank yielded comparable results. MR confirmed most causal pathways predicted by the Bayesian networks.Here, BasalISR had the highest causal effect on fatty liver predisposition, providing mechanistic evidence underpinning the established association of NAFLD and T2D. BasalISR may represent a pragmatic biomarker for NAFLD prediction in clinical practice.

Background Type 2 diabetes (T2D) is common and increasing in prevalence. It is possible to prevent or delay T2D using lifestyle intervention programmes. Entry to these programmes is usually determined by a measure of glycaemia in the ‘intermediate’ range. This paper investigated the relationship between HbA1c and future diabetes risk and determined the impact of varying thresholds to identify those at high risk of developing T2D. Methods We studied 4227 participants without diabetes aged ≥ 40 years recruited to the Exeter 10,000 population cohort in South West England. HbA1c was measured at study recruitment with repeat HbA1c available as part of usual care. Absolute risk of developing diabetes within 5 years, defined by HbA1c ≥ 48 mmol/mol (6.5%), according to baseline HbA1c, was assessed by a flexible parametric survival model. Results The overall absolute 5-year risk (95% CI) of developing T2D in the cohort was 4.2% (3.6, 4.8%). This rose to 7.1% (6.1, 8.2%) in the 56% (n = 2358/4224) of participants classified ‘high-risk’ with HbA1c ≥ 39 mmol/mol (5.7%; ADA criteria). Under IEC criteria, HbA1c ≥ 42 mmol/mol (6.0%), 22% (n = 929/4277) of the cohort was classified high-risk with 5-year risk 14.9% (12.6, 17.2%). Those with the highest HbA1c values (44–47 mmol/mol [6.2–6.4%]) had much higher 5-year risk, 26.4% (22.0, 30.5%) compared with 2.1% (1.5, 2.6%) for 39–41 mmol/mol (5.7–5.9%) and 7.0% (5.4, 8.6%) for 42–43 mmol/mol (6.0–6.1%). Changing the entry criterion to prevention programmes from 39 to 42 mmol/mol (5.7–6.0%) reduced the proportion classified high-risk by 61%, and increased the positive predictive value (PPV) from 5.8 to 12.4% with negligible impact on the negative predictive value (NPV), 99.6% to 99.1%. Increasing the threshold further, to 44 mmol/mol (6.2%), reduced those classified high-risk by 59%, and markedly increased the PPV from 12.4 to 23.2% and had little impact on the NPV (99.1% to 98.5%). Conclusions A large proportion of people are identified as high-risk using current thresholds. Increasing the risk threshold markedly reduces the number of people that would be classified as high-risk and entered into prevention programmes, although this must be balanced against cases missed. Raising the entry threshold would allow limited intervention opportunities to be focused on those most likely to develop T2D.

IntroductionZinc transporter 8 autoantibodies (ZnT8A) are biomarkers of beta cell autoimmunity in type 1 diabetes that have become more widely available to clinicians in recent years. Robust control population-defined thresholds are essential to ensure high clinical specificity in islet autoantibody testing. We aimed to determine the optimal cut-offs for ZnT8A testing.Research design and methods97.5th and 99th centile cut-offs were determined using residual clinical sera from 1559 controls aged between 0 and 83 years with no history of diabetes and a hemoglobin A1c level of less than 6.0% (ResultsUsing the manufacturer’s limit of detection, 6 WHO U/mL, 16.2% of people in the control cohort had detectable levels of ZnT8A and those who had detectable ZnT8A were much more likely to be younger (pConclusionsAge-related thresholds are needed for ZnT8A testing. In those aged

OBJECTIVE Gastrointestinal adverse effects occur in 20–30% of patients with metformin-treated type 2 diabetes, leading to premature discontinuation in 5–10% of the cases. Gastrointestinal intolerance may reflect localized high concentrations of metformin in the gut. We hypothesized that reduced transport of metformin via the plasma membrane monoamine transporter (PMAT) and organic cation transporter 1 (OCT1) could increase the risk of severe gastrointestinal adverse effects. RESEARCH DESIGN AND METHODS The study included 286 severe metformin-intolerant and 1,128 metformin-tolerant individuals from the IMI DIRECT (Innovative Medicines Initiative: DIabetes REsearCh on patient straTification) consortium. We assessed the association of patient characteristics, concomitant medication, and the burden of mutations in the SLC29A4 and SLC22A1 genes on odds of intolerance. RESULTS Women (P < 0.001) and older people (P < 0.001) were more likely to develop metformin intolerance. Concomitant use of transporter-inhibiting drugs increased the odds of intolerance (odds ratio [OR] 1.72, P < 0.001). In an adjusted logistic regression model, the G allele at rs3889348 (SLC29A4) was associated with gastrointestinal intolerance (OR 1.34, P = 0.005). rs3889348 is the top cis-expression quantitative trait locus for SLC29A4 in gut tissue where carriers of the G allele had reduced expression. Homozygous carriers of the G allele treated with transporter-inhibiting drugs had more than three times higher odds of intolerance compared with carriers of no G allele and not treated with inhibiting drugs (OR 3.23, P < 0.001). Use of a genetic risk score derived from rs3889348 and SLC22A1 variants found that the odds of intolerance were more than twice as high in individuals who carry three or more risk alleles compared with those carrying none (OR 2.15, P = 0.01). CONCLUSIONS These results suggest that intestinal metformin transporters and concomitant medications play an important role in the gastrointestinal adverse effects of metformin.

Linear mixed effects models are frequently used in biomedical statistics to model the trajectory of a repeatedly measured longitudinal variable, such as a biomarker, over time. However, population-level estimates may be biased by censoring bias resulting from exit criteria that depend on the variable in question. A joint longitudinal-survival model, in which the exit criteria and longitudinal variable are modelled simultaneously, may address this bias. Using blood glucose progression (change in HbA1c) in type 2 diabetes patients on metformin monotherapy as an example, we study the potential benefit of using joint models to model trajectory of a biomarker in observational data. 7,712 patients with type 2 diabetes initiating metformin monotherapy were identified in UK Biobank’s general practice (GP) linked records. Genetic information was extracted from UK Biobank, and prescription records, baseline clinical features and biomarkers, and longitudinal HbA1c measures were extracted from GP records. Exit criteria for follow-up for a patient was defined as progression to an additional glucose-lowering drug (which is more likely in patient with higher HbA1c). Estimates of HbA1c trajectory over time were compared using linear mixed effect model approaches (which do not account for censoring bias) and joint models. In the primary analysis, a 0.19 mmol/mol per year higher (p = 0.01) HbA1c gradient was estimated using the joint model compared to the linear mixed effects model. This difference between models was attenuated (0.13 mmol/mol per year higher, p=0.43) when baseline clinical features and biomarkers were included as additional covariates.Censoring bias should be carefully considered when modelling trajectories of repeatedly measured longitudinal variables in observational data. Joint longitudinal-survival models are a useful approach to identify and potentially correct for censoring bias when estimating population-level trajectories.Author SummaryModelling biomarkers that change over time in real world data is a challenging statistical problem due to many potential sources of bias. For example, when studying a chronic disease using a biomarker or other measurement that represents disease severity, medication intended to affect that measurement has a profound effect on how it will change over time. One common way to control for this is to study a cohort on the same treatment strategy. That way, results are not influenced by treatment change. If a patient progresses to stronger medication, then future data is no longer used. However, this approach introduces its own bias. Patients whose condition progress particularly quickly are more likely to change treatment more rapidly (and therefore be removed from further analysis, or ‘censored’), so the cohort is biased towards those whose condition progresses slower. In this paper we apply a technique called joint longitudinal-survival modelling which can adjust for this censoring bias and produce less biased estimates of progression rates. We use HbA1c (a widely used measure of glucose control) in type 2 diabetes as an example, however our methods are theoretically applicable to a range of problems across medicine in which a biomarker or feature is repeatedly measured in an individual.

Latent autoimmune diabetes of adults (LADA) is typically defined as a new diabetes diagnosis after 35 years of age, presenting with clinical features of type 2 diabetes, in whom a type 1 diabetes–associated islet autoantibody is detected. Identifying autoimmune diabetes is important since the prognosis and optimal therapy differ. However, the existing LADA definition identifies a group with clinical and genetic features intermediate between typical type 1 and type 2 diabetes. It is unclear whether this is due to 1) true autoimmune diabetes with a milder phenotype at older onset ages that initially appears similar to type 2 diabetes but later requires insulin, 2) a disease syndrome where the pathophysiologies of type 1 and type 2 diabetes are both present in each patient, or 3) a heterogeneous group resulting from difficulties in classification. Herein, we suggest that difficulties in classification are a major component resulting from defining LADA using a diagnostic test—islet autoantibody measurement—with imperfect specificity applied in low-prevalence populations. This yields a heterogeneous group of true positives (autoimmune type 1 diabetes) and false positives (nonautoimmune type 2 diabetes). For clinicians, this means that islet autoantibody testing should not be undertaken in patients who do not have clinical features suggestive of autoimmune diabetes: in an adult without clinical features of type 1 diabetes, it is likely that a single positive antibody will represent a false-positive result. This is in contrast to patients with features suggestive of type 1 diabetes, where false-positive results will be rare. For researchers, this means that current definitions of LADA are not appropriate for the study of autoimmune diabetes in later life. Approaches that increase test specificity, or prior likelihood of autoimmune diabetes, are needed to avoid inclusion of participants who have nonautoimmune (type 2) diabetes. Improved classification will allow improved assignment of prognosis and therapy as well as an improved cohort in which to analyze and better understand the detailed pathophysiological components acting at onset and during disease progression in late-onset autoimmune diabetes.

Introduction Low high-density lipoprotein (HDL) is widely used as a marker of cardiovascular disease risk, although this relationship is not causal and is likely mediated through associations with other risk factors. Low HDL is extremely common in sub-Saharan African populations, and this has often been interpreted to indicate that these populations will have increased cardiovascular risk. We aimed to determine whether the association between HDL and other cardiovascular risk factors differed between populations in sub-Saharan Africa and the UK.Methods We compared data from adults living in Uganda and Malawi (n=26 216) and in the UK (n=8747). We examined unadjusted and adjusted levels of HDL and applied the WHO recommended cut-offs for prevalence estimates. We used spline and linear regression to assess the relationship between HDL and other cardiovascular risk factors.Results HDL was substantially lower in the African than in the European studies (geometric mean 0.9textendash1.2 mmol/L vs 1.3textendash1.8 mmol/L), with African prevalence of low HDL as high as 77geometric mean 3.3textendash3.9 mmol/L vs 4.6textendash5.4 mmol/L). In comparison with European studies the relationship between HDL and adiposity (body mass index, waist to hip ratio) was greatly attenuated in African studies and the relationship with non-HDL cholesterol reversed: in African studies low HDL was associated with lower non-HDL cholesterol. The association between sex and HDL was also different; using the WHO sex-specific definitions, low HDL was substantially more common among women (69textendash77 than men (41textendash59 in Uganda/Malawi.Conclusion The relationship between HDL and sex, adiposity and non-HDL cholesterol in sub-Saharan Africa is different from European populations. In sub-Saharan Africans low HDL is a marker of low overall cholesterol and sex differences are markedly attenuated. Therefore low HDL in isolation is unlikely to indicate raised cardiovascular risk and the WHO sex-based cut-offs are inappropriate.Data from the EXTEND study are available through application to the Peninsula Research Bank (https://exetercrfnihr.org/about/exeter-10000-prb/). For enquiries about access to the Exeter Family Study please contact BS (B.Shields@exeter.ac.uk). The MEIRU data are available through the LSHTM Research Data Compass (https://datacompass.lshtm.ac.uk/961/).

Recent epidemiological data have shown that more than half of all new cases of type 1 diabetes occur in adults. Key genetic, immune, and metabolic differences exist between adult- and childhood-onset type 1 diabetes, many of which are not well understood. A substantial risk of misclassification of diabetes type can result. Notably, some adults with type 1 diabetes may not require insulin at diagnosis, their clinical disease can masquerade as type 2 diabetes, and the consequent misclassification may result in inappropriate treatment. In response to this important issue, JDRF convened a workshop of international experts in November 2019. Here, we summarize the current understanding and unanswered questions in the field based on those discussions, highlighting epidemiology and immunogenetic and metabolic characteristics of adult-onset type 1 diabetes as well as disease-associated comorbidities and psychosocial challenges. In adult-onset, as compared with childhood-onset, type 1 diabetes, HLA-associated risk is lower, with more protective genotypes and lower genetic risk scores; multiple diabetes-associated autoantibodies are decreased, though GADA remains dominant. Before diagnosis, those with autoantibodies progress more slowly, and at diagnosis, serum C-peptide is higher in adults than children, with ketoacidosis being less frequent. Tools to distinguish types of diabetes are discussed, including body phenotype, clinical course, family history, autoantibodies, comorbidities, and C-peptide. By providing this perspective, we aim to improve the management of adults presenting with type 1 diabetes.

Genetic variants identified by genome-wide association studies can contribute to disease risk by altering the production and abundance of mRNA, proteins and other molecules. However, the interplay between molecular intermediaries that define the pathway from genetic variation to disease is not well understood. Here, we evaluated the shared genetic regulation of mRNA molecules, proteins and metabolites derived from whole blood from 3,029 human donors. We find abundant allelic heterogeneity, where multiple variants regulate a particular molecular phenotype, and pleiotropy, where a single variant was associated with multiple molecular phenotypes over multiple genomic regions. We find varying proportions of shared genetic regulation across phenotypes, highest between expression and proteins (66.6%). We were able to recapitulate a substantial proportion of gene expression genetic regulation in a diverse set of 44 tissues, with a median of 88% shared associations for blood expression and 22.3% for plasma proteins. Finally, the genetic and molecular associations were represented in networks including 2,828 known GWAS variants. One sub-network shows the trans relationship between rs149007767 and RTEN, and identifies GRB10 and IKZF1 as candidates mediating genes. Our work provides a roadmap to understanding molecular networks and deriving the underlying mechanism of action of GWAS variants across different molecular phenotypes.

OBJECTIVE We investigated the processes underlying glycemic deterioration in type 2 diabetes (T2D). RESEARCH DESIGN AND METHODS A total of 732 recently diagnosed patients with T2D from the Innovative Medicines Initiative Diabetes Research on Patient Stratification (IMI DIRECT) study were extensively phenotyped over 3 years, including measures of insulin sensitivity (OGIS), β-cell glucose sensitivity (GS), and insulin clearance (CLIm) from mixed meal tests, liver enzymes, lipid profiles, and baseline regional fat from MRI. The associations between the longitudinal metabolic patterns and HbA1c deterioration, adjusted for changes in BMI and in diabetes medications, were assessed via stepwise multivariable linear and logistic regression. RESULTS Faster HbA1c progression was independently associated with faster deterioration of OGIS and GS and increasing CLIm; visceral or liver fat, HDL-cholesterol, and triglycerides had further independent, though weaker, roles (R2 = 0.38). A subgroup of patients with a markedly higher progression rate (fast progressors) was clearly distinguishable considering these variables only (discrimination capacity from area under the receiver operating characteristic = 0.94). The proportion of fast progressors was reduced from 56% to 8–10% in subgroups in which only one trait among OGIS, GS, and CLIm was relatively stable (odds ratios 0.07–0.09). T2D polygenic risk score and baseline pancreatic fat, glucagon-like peptide 1, glucagon, diet, and physical activity did not show an independent role. CONCLUSIONS Deteriorating insulin sensitivity and β-cell function, increasing insulin clearance, high visceral or liver fat, and worsening of the lipid profile are the crucial factors mediating glycemic deterioration of patients with T2D in the initial phase of the disease. Stabilization of a single trait among insulin sensitivity, β-cell function, and insulin clearance may be relevant to prevent progression.

Context Pancreatic beta-cell glucose sensitivity is the slope of the plasma glucose-insulin secretion relationship and is a key predictor of deteriorating glucose tolerance and development of type 2 diabetes. However, there are no large-scale studies looking at the genetic determinants of beta-cell glucose sensitivity. Objective To understand the genetic determinants of pancreatic beta-cell glucose sensitivity using genome-wide meta-analysis and candidate gene studies. Design We performed a genome-wide meta-analysis for beta-cell glucose sensitivity in subjects with type 2 diabetes and nondiabetic subjects from 6 independent cohorts (n = 5706). Beta-cell glucose sensitivity was calculated from mixed meal and oral glucose tolerance tests, and its associations between known glycemia-related single nucleotide polymorphisms (SNPs) and genome-wide association study (GWAS) SNPs were estimated using linear regression models. Results Beta-cell glucose sensitivity was moderately heritable (h2 ranged from 34% to 55%) using SNP and family-based analyses. GWAS meta-analysis identified multiple correlated SNPs in the CDKAL1 gene and GIPR-QPCTL gene loci that reached genome-wide significance, with SNP rs2238691 in GIPR-QPCTL (P value = 2.64 × 10−9) and rs9368219 in the CDKAL1 (P value = 3.15 × 10−9) showing the strongest association with beta-cell glucose sensitivity. These loci surpassed genome-wide significance when the GWAS meta-analysis was repeated after exclusion of the diabetic subjects. After correction for multiple testing, glycemia-associated SNPs in or near the HHEX and IGF2B2 loci were also associated with beta-cell glucose sensitivity. Conclusion We show that, variation at the GIPR-QPCTL and CDKAL1 loci are key determinants of pancreatic beta-cell glucose sensitivity.

IntroductionPharmaceutical treatment options for patients with type 2 diabetes mellitus (T2DM) have increased to include multiple classes of oral glucose-lowering agents but without accompanying guidance on which of these may most benefit individual patients. Clinicians lack information for treatment intensification after first-line metformin therapy. Stratifying patients by simple clinical characteristics may improve care by targeting treatment options to those in whom they are most effective. This academically designed and run three-way crossover trial aims to test a stratification approach using three standard oral glucose-lowering agents.Methods and analysisTriMaster is a randomised, double-blind, crossover trial taking place at up to 25 clinical sites across England, Scotland and Wales. 520 patients with T2DM treated with either metformin alone, or metformin and a sulfonylurea who have glycated haemoglobin (HbA1c) >58 mmol/mol will be randomised to receive 16 weeks each of a dipeptidyl peptidase‐4 inhibitor, sodium-glucose co-transporter-2 inhibitor and thiazolidinedione in random order. Participants will be assessed at the end of each treatment period, providing clinical and biochemical data, and their experience of side effects. Participant preference will be assessed on completion of all three treatments. The primary endpoint is HbA1c after 4 months of therapy (allowing a range of 12–18 weeks for analysis). Secondary endpoints include participant-reported preference between the three treatments, tolerability and prevalence of side effects.Ethical approvalThis study was approved by National Health Service Health Research Authority Research Ethics Committee South Central—Oxford A, study 16/SC/0147. Written informed consent will be obtained from all participants. Results will be submitted to a peer-reviewed journal and presented at relevant scientific meetings. A lay summary of results will be made available to all participants.Trial registration numbers12039221; 2015-002790-38 and NCT02653209.

BackgroundThe rising prevalence of type 2 diabetes (T2D) poses a major global challenge. It remains unresolved to what extent transcriptomic signatures of metabolic dysregulation and T2D can be observed in easily accessible tissues such as blood. Additionally, large-scale human studies are required to further our understanding of the putative inflammatory component of insulin resistance and T2D. Here we used transcriptomics data from individuals with (n = 789) and without (n = 2127) T2D from the IMI-DIRECT cohorts to describe the co-expression structure of whole blood that mainly reflects processes and cell types of the immune system, and how it relates to metabolically relevant clinical traits and T2D.MethodsClusters of co-expressed genes were identified in the non-diabetic IMI-DIRECT cohort and evaluated with regard to stability, as well as preservation and rewiring in the cohort of individuals with T2D. We performed functional and immune cell signature enrichment analyses, and a genome-wide association study to describe the genetic regulation of the modules. Phenotypic and trans-omics associations of the transcriptomic modules were investigated across both IMI-DIRECT cohorts.ResultsWe identified 55 whole blood co-expression modules, some of which clustered in larger super-modules. We identified a large number of associations between these transcriptomic modules and measures of insulin action and glucose tolerance. Some of the metabolically linked modules reflect neutrophil-lymphocyte ratio in blood while others are independent of white blood cell estimates, including a module of genes encoding neutrophil granule proteins with antibacterial properties for which the strongest associations with clinical traits and T2D status were observed. Through the integration of genetic and multi-omics data, we provide a holistic view of the regulation and molecular context of whole blood transcriptomic modules. We furthermore identified an overlap between genetic signals for T2D and co-expression modules involved in type II interferon signaling.ConclusionsOur results offer a large-scale map of whole blood transcriptomic modules in the context of metabolic disease and point to novel biological candidates for future studies related to T2D.

Background: Dietary advice remains the cornerstone of prevention and management of type 2 diabetes (T2D). However, understanding the efficacy of dietary interventions is confounded by the challenges inherent in assessing free living diet. Here we profiled dietary metabolites to investigate glycaemic deterioration and cardiometabolic risk in people at risk of or living with T2D. Methods: We analysed data from plasma collected at baseline and 18-month follow-up in individuals from the Innovative Medicines Initiative (IMI) Diabetes Research on Patient Stratification (DIRECT) cohort 1 n = 403 individuals with normal or impaired glucose regulation (prediabetic) and cohort 2 n = 458 individuals with new onset of T2D. A dietary metabolite profile model (Tpred) was constructed using multivariable regression of 113 plasma metabolites obtained from targeted metabolomics assays. The continuous Tpred score was used to explore the relationships between diet, glycaemic deterioration and cardio-metabolic risk via multiple linear regression models. Findings: A higher Tpred score was associated with healthier diets high in wholegrain (β=3.36 g, 95% CI 0.31, 6.40 and β=2.82 g, 95% CI 0.06, 5.57) and lower energy intake (β=-75.53 kcal, 95% CI -144.71, -2.35 and β=-122.51 kcal, 95% CI -186.56, -38.46), and saturated fat (β=-0.92 g, 95% CI -1.56, -0.28 and β=–0.98 g, 95% CI -1.53, -0.42 g), respectively for cohort 1 and 2. In both cohorts a higher Tpred score was also associated with lower total body adiposity and favourable lipid profiles HDL-cholesterol (β=0.07 mmol/L, 95% CI 0.03, 0.1), (β=0.08 mmol/L, 95% CI 0.04, 0.1), and triglycerides (β=-0.1 mmol/L, 95% CI -0.2, -0.03), (β=-0.2 mmol/L, 95% CI -0.3, -0.09), respectively for cohort 1 and 2. In cohort 2, the Tpred score was negatively associated with liver fat (β=-0.74%, 95% CI -0.67, -0.81), and lower fasting concentrations of HbA1c (β=-0.9 mmol/mol, 95% CI -1.5, -0.1), glucose (β=-0.2 mmol/L, 95% CI -0.4, -0.05) and insulin (β=-11.0 pmol/mol, 95% CI -19.5, -2.6). Longitudinal analysis showed at 18-month follow up a higher Tpred score was also associated lower total body adiposity in both cohorts and lower fasting glucose (β=-0.2 mmol/L, 95% CI -0.3, -0.01) and insulin (β=-9.2 pmol/mol, 95% CI -17.9, -0.4) concentrations in cohort 2. Interpretation: Plasma dietary metabolite profiling provides objective measures of diet intake, showing a relationship to glycaemic deterioration and cardiometabolic health. Funding: This work was supported by the Innovative Medicines Initiative Joint Undertaking under grant agreement no. 115,317 (DIRECT), resources of which are composed of financial contribution from the European Union's Seventh Framework Programme (FP7/2007–2013) and EFPIA companies.

AIMS Misclassification of diabetes is common due to an overlap in the clinical features of type 1 and type 2 diabetes. Combined diagnostic models incorporating clinical and biomarker information have recently been developed that can aid classification, but they have not been validated using pancreatic pathology. We evaluated a clinical diagnostic model against histologically defined type 1 diabetes. METHODS We classified cases from the Network for Pancreatic Organ donors with Diabetes (nPOD) biobank as type 1 (n = 111) or non-type 1 (n = 42) diabetes using histopathology. Type 1 diabetes was defined by lobular loss of insulin-containing islets along with multiple insulin-deficient islets. We assessed the discriminative performance of previously described type 1 diabetes diagnostic models, based on clinical features (age at diagnosis, BMI) and biomarker data [autoantibodies, type 1 diabetes genetic risk score (T1D-GRS)], and singular features for identifying type 1 diabetes by the area under the curve of the receiver operator characteristic (AUC-ROC). RESULTS Diagnostic models validated well against histologically defined type 1 diabetes. The model combining clinical features, islet autoantibodies and T1D-GRS was strongly discriminative of type 1 diabetes, and performed better than clinical features alone (AUC-ROC 0.97 vs. 0.95; P = 0.03). Histological classification of type 1 diabetes was concordant with serum C-peptide [median

BackgroundNon-alcoholic fatty liver disease (NAFLD) is highly prevalent and causes serious health complications in type 2 diabetes (T2D) and beyond. Early diagnosis of NAFLD is important, as this can help prevent irreversible damage to the liver and ultimately hepatocellular carcinomas.Methods and FindingsUtilizing the baseline data from the IMI DIRECT participants (n=1514) we sought to expand etiological understanding and develop a diagnostic tool for NAFLD using machine learning. Multi-omic (genetic, transcriptomic, proteomic, and metabolomic) and clinical (liver enzymes and other serological biomarkers, anthropometry, and measures of beta-cell function, insulin sensitivity, and lifestyle) data comprised the key input variables. The models were trained on MRI image-derived liver fat content (ConclusionsWe have developed clinically useful liver fat prediction models (see:www.predictliverfat.org) and identified biological features that appear to affect liver fat accumulation.

OBJECTIVE A precision approach to type 2 diabetes therapy would aim to target treatment according to patient characteristics. We examined if measures of insulin resistance and secretion were associated with glycemic response to dipeptidyl peptidase 4 (DPP-4) inhibitor therapy. RESEARCH DESIGN AND METHODS We evaluated whether markers of insulin resistance and insulin secretion were associated with 6-month glycemic response in a prospective study of noninsulin-treated participants starting DPP-4 inhibitor therapy (Predicting Response to Incretin Based Agents [PRIBA] study; n = 254), with replication for routinely available markers in U.K. electronic health care records (Clinical Practice Research Datalink [CPRD]; n = 23,001). In CPRD, we evaluated associations between baseline markers and 3-year durability of response. To test the specificity of findings, we repeated analyses for glucagon-like peptide 1 (GLP-1) receptor agonists (PRIBA, n = 339; CPRD, n = 4,464). RESULTS In PRIBA, markers of higher insulin resistance (higher fasting C-peptide [P = 0.03], HOMA2 insulin resistance [P = 0.01], and triglycerides [P < 0.01]) were associated with reduced 6-month HbA1c response to DPP-4 inhibitors. In CPRD, higher triglycerides and BMI were associated with reduced HbA1c response (both P < 0.01). A subgroup defined by obesity (BMI ≥30 kg/m2) and high triglycerides (≥2.3 mmol/L) had reduced 6-month response in both data sets (PRIBA HbA1c reduction 5.3 [95% CI 1.8, 8.6] mmol/mol [0.5%] [obese and high triglycerides] vs. 11.3 [8.4, 14.1] mmol/mol [1.0%] [nonobese and normal triglycerides]; P = 0.01). In CPRD, the obese, high- triglycerides subgroup also had less durable response (hazard ratio 1.28 [1.16, 1.41]; P < 0.001). There was no association between markers of insulin resistance and response to GLP-1 receptor agonists. CONCLUSIONS Markers of higher insulin resistance are consistently associated with reduced glycemic response to DPP-4 inhibitors. This finding provides a starting point for the application of a precision diabetes approach to DPP-4 inhibitor therapy.

Samples for adrenocorticotropic hormone (ACTH) and aldosterone/renin analysis usually require rapid transport to the receiving laboratory for immediate separation and freezing. In practice, this means assessment is limited to hospital settings and many samples are rejected. We examined whether these requirements are necessary by assessing the stability of ACTH, aldosterone and renin over 48 hours in whole blood collected in serum gel and EDTA plasma from 31 participants. Our results show that ACTH collected into EDTA plasma is stable at room temperature for at least 6 hours, mean change at 6 hours –2.6% (95% CI –9.7 to 4.5). Both aldosterone and renin were stable collected on serum gel at room temperature for at least 6 hours: mean change aldosterone +0.2% (95% CI –3.6 to 4.0), renin –1.9% (95% CI –7.0 to3.2). Therefore, by using appropriate preservatives, ACTH and aldosterone/renin can be measured on samples collected at room temperature and processed within 6 hours. This would facilitate outpatient and emergency room assessment of these analytes.

Introduction: Urinary C-peptide creatinine ratio (UCPCR) measured in urine collected at home after a meal has been shown to correlate strongly with stimulated blood C-peptide in European populations. This association and the clinical utility of UCPCR in a sub-Saharan African clinical setting has not been described before. We aimed to assess the performance of UCPCR as a measure of endogenous insulin secretion in Cameroon.Methods: UCPCR was measured on two separate days before and after a standard mixed-meal tolerance test (MMTT), after lunch and after supper in 14 patients with diabetes and 14 healthy control individuals. Blood C-peptide was measured serially during a standard 75g oral glucose tolerance test (OGTT) every 30 minutes, on a separate day. The primary outcome was the correlation between stimulated blood C-peptide levels and post-meal UCPCR values.Results: stimulated blood C-peptide was significantly lower in participants with diabetes vs controls; median (IQR) 719 (110-999) pmol/l vs 1080 (934-1820) pmol/l, p=0.04. Fasting and post-MMTT UCPCR correlated strongly with stimulated blood C-peptide measurement in participants with diabetes (r= 0.71, p= 0.005) and (r=0.71, p=0.004) respectively. Post-meal UCPCR showed a moderate or poor correlation with stimulated blood C-peptide levels (supper r=0.56, p=0.04, lunch r=0.31, p=0.29). In participants without diabetes, UCPCR showed no relationship with stimulated blood C-peptide (supper r= -0.03, p=0.92, lunch r= -0.06, p=0.83). Conclusion: these results suggest that post-meal UCPCR performs less well in this population compared to European populations and may not be useful to assess endogenous insulin secretion in participants without diabetes in Cameroon.

Background: Low high-density lipoprotein (HDL) is widely used as a marker of cardiovascular disease risk, despite lack of evidence of a casual role. It is extremely common in sub-Saharan African populations, where reported cardiovascular risk is low. We aimed to determine whether the association between HDL and other cardiovascular risk factors differed between populations in sub-Saharan Africa and the UK. Methods: We compared data from adults living in Uganda and Malawi (n=26,216) and the UK (n=8,747). We examined unadjusted and adjusted levels of HDL, and applied WHO recommended cut-offs for prevalence estimates. We used spline and linear regression to assess the relationship between HDL and other cardiovascular risk factors. Findings: Geometric-mean HDL was substantially lower in the African than European studies (0.9-1.2mmol/l vs 1.3-1.8mmol/l), with African prevalence of low-HDL as high as 77%. In comparison to European studies the relationship between HDL and adiposity (BMI, waist-hip ratio) was greatly attenuated in African studies, and the relationship with non-HDL cholesterol reversed: low HDL was associated with lower non-HDL. The association between sex and HDL was also markedly different; using WHO sex-specific definitions low-HDL was substantially more common among women (69-77%) than men (41-59%) in Uganda/Malawi. Interpretation: The relationship between HDL and sex, adiposity and Non-HDL cholesterol in sub-Saharan Africa is markedly different from European populations. Low HDL is a marker of low overall cholesterol, and sex differences are markedly attenuated. Therefore low HDL may not be a reliable marker of cardiovascular risk in Sub-Saharan Africa, and WHO sex based cut-offs are inappropriate. Funding Statement: This work is funded by the National Institute of Health and Social Care (UK), NIHR Global Health Group 17/63/131. Declaration of Interests: All authors report no conflicts of interest in relation to this work. Ethics Approval Statement: The Ugandan cohort was approved by the Uganda Virus Research Institute Science and Ethics Committee and the Uganda National Council of Science and Technology. The Malawian study was approved by the National Health Sciences Research Committee of Malawi nd London School of Hygiene and Tropical Medicine Ethics Committee, UK.The Exeter 10,000 study was approved by the UK NHS Health Research Authority South West Ethics Committee. The Exeter Family Study was approved by the University of Exeter Research Ethics Committee, UK.

Background There is much interest in the use of prognostic and diagnostic prediction models in all areas of clinical medicine. The use of machine learning to improve prognostic and diagnostic accuracy in this area has been increasing at the expense of classic statistical models. Previous studies have compared performance between these two approaches but their findings are inconsistent and many have limitations. We aimed to compare the discrimination and calibration of seven models built using logistic regression and optimised machine learning algorithms in a clinical setting, where the number of potential predictors is often limited, and externally validate the models. Methods We trained models using logistic regression and six commonly used machine learning algorithms to predict if a patient diagnosed with diabetes has type 1 diabetes (versus type 2 diabetes). We used seven predictor variables (age, BMI, GADA islet-autoantibodies, sex, total cholesterol, HDL cholesterol and triglyceride) using a UK cohort of adult participants (aged 18–50 years) with clinically diagnosed diabetes recruited from primary and secondary care (n = 960, 14% with type 1 diabetes). Discrimination performance (ROC AUC), calibration and decision curve analysis of each approach was compared in a separate external validation dataset (n = 504, 21% with type 1 diabetes). Results Average performance obtained in internal validation was similar in all models (ROC AUC ≥ 0.94). In external validation, there were very modest reductions in discrimination with AUC ROC remaining ≥ 0.93 for all methods. Logistic regression had the numerically highest value in external validation (ROC AUC 0.95). Logistic regression had good performance in terms of calibration and decision curve analysis. Neural network and gradient boosting machine had the best calibration performance. Both logistic regression and support vector machine had good decision curve analysis for clinical useful threshold probabilities. Conclusion Logistic regression performed as well as optimised machine algorithms to classify patients with type 1 and type 2 diabetes. This study highlights the utility of comparing traditional regression modelling to machine learning, particularly when using a small number of well understood, strong predictor variables.