First-in-human, double-blind, randomized phase 1b study of peptide immunotherapy IMCY-0098 in new-onset type 1 diabetes.

BACKGROUND: Type 1 diabetes (T1D) is a CD4+ T cell-driven autoimmune disease characterized by the destruction of insulin-producing pancreatic ß-cells by CD8+ T cells. Achieving glycemic targets in T1D remains challenging in clinical practice; new treatments aim to halt autoimmunity and prolong ß-cell survival. IMCY-0098 is a peptide derived from human proinsulin that contains a thiol-disulfide oxidoreductase motif at the N-terminus and was developed to halt disease progression by promoting the specific elimination of pathogenic T cells. METHODS: This first-in-human, 24-week, double-blind phase 1b study evaluated the safety of three dosages of IMCY-0098 in adults diagnosed with T1D < 6 months before study start. Forty-one participants were randomized to receive four bi-weekly injections of placebo or increasing doses of IMCY-0098 (dose groups A/B/C received 50/150/450 µg for priming followed by three further administrations of 25/75/225 µg, respectively). Multiple T1D-related clinical parameters were also assessed to monitor disease progression and inform future development. Long-term follow-up to 48 weeks was also conducted in a subset of patients. RESULTS: Treatment with IMCY-0098 was well tolerated with no systemic reactions; a total of 315 adverse events (AEs) were reported in 40 patients (97.6%) and were related to study treatment in 29 patients (68.3%). AEs were generally mild; no AE led to discontinuation of the study or death. No significant decline in C-peptide was noted from baseline to Week 24 for dose A, B, C, or placebo (mean change - 0.108, - 0.041, - 0.040, and - 0.012, respectively), suggesting no disease progression. CONCLUSIONS: Promising safety profile and preliminary clinical response data support the design of a phase 2 study of IMCY-0098 in patients with recent-onset T1D. TRIAL REGISTRATION: IMCY-T1D-001: ClinicalTrials.gov NCT03272269; EudraCT: 2016-003514-27; and IMCY-T1D-002: ClinicalTrials.gov NCT04190693; EudraCT: 2018-003728-35.

Development of an exoglycosidase plate-based assay for detecting α1-3,4 fucosylation biomarker in individuals with HNF1A-MODY.

Maturity-onset diabetes of the young due to hepatocyte nuclear factor-1 alpha variants (HNF1A-MODY) causes monogenic diabetes. Individuals carrying damaging variants in HNF1A show decreased levels of α1-3,4 fucosylation, as demonstrated on antennary fucosylation of blood plasma N-glycans. The excellent diagnostic performance of this glycan biomarker in blood plasma N-glycans of individuals with HNF1A-MODY has been demonstrated using liquid chromatography methods. Here, we have developed a high-throughput exoglycosidase plate-based assay to measure α1-3,4 fucosylation levels in blood plasma samples. The assay has been optimized and its validity tested using 1000 clinical samples from a cohort of individuals with young-adult onset diabetes including cases with HNF1A-MODY. The α1-3,4 fucosylation levels in blood plasma showed a good differentiating power in identifying cases with damaging HNF1A variants, as demonstrated by receiver operating characteristic curve analysis with the AUC values of 0.87 and 0.95. This study supports future development of a simple diagnostic test to measure this glycan biomarker for application in a clinical setting.

Interlaboratory evaluation of plasma N-glycan antennary fucosylation as a clinical biomarker for HNF1A-MODY using liquid chromatography methods.

Antennary fucosylation alterations in plasma glycoproteins have been previously proposed and tested as a biomarker for differentiation of maturity onset diabetes of the young (MODY) patients carrying a functional mutation in the HNF1A gene. Here, we developed a novel LC-based workflow to analyze blood plasma N-glycan fucosylation in 320 diabetes cases with clinical features matching those at risk of HNF1A-MODY. Fucosylation levels measured in two independent research centers by using similar LC-based methods were correlated to evaluate the interlaboratory performance of the biomarker. The interlaboratory study showed good correlation between fucosylation levels measured for the 320 cases in the two centers with the correlation coefficient (r) of up to 0.88 for a single trait A3FG3S2. The improved chromatographic separation allowed the identification of six single glycan traits and a derived antennary fucosylation trait that were able to differentiate individuals carrying pathogenic mutations from benign or no HNF1A mutation cases, as determined by the area under the curve (AUC) of up to 0.94. The excellent (r = 0.88) interlaboratory performance of the glycan biomarker for HNF1A-MODY further supports the development of a clinically relevant diagnostic test measuring antennary fucosylation levels.

Altered cortisol metabolism in individuals with HNF1A-MODY.

OBJECTIVE AND CONTEXT: Maturity onset diabetes of the young due to variants in HNF1A (HNF1A-MODY) is the most common form of monogenic diabetes. Individuals with HNF1A-MODY usually have a lean phenotype which contrasts with type 2 diabetes (T2DM). Data from hepatocytes derived from Hnf1a knock-out mice demonstrated dysregulation of 11ß-hydroxysteroid dehydrogenase type 1 (11ß-HSD1), which regulates glucocorticoid availability and action in target tissues, together with 11ß-HSD2 and steroid A-ring reductases, 5α- and 5ß-reductase. We proposed that altered glucocorticoid metabolism might underpin some of the phenotypic differences between patients with HNF1A-MODY and those with T2DM. DESIGN: A retrospective matched cohort study. PATIENTS AND MEASUREMENTS: 24-hours urine steroid metabolome profiling was carried out by gas chromatography-mass spectrometry in 35 subjects with HNF1A-MODY, 35 individuals with T2DM and 35 healthy controls matched for age, sex and BMI. The steroid metabolites were expressed as µg/L in all groups and measured in mid-morning urine in diabetic subjects and 24-hour urine collection in healthy controls. Hence, only ratios were compared not the individual steroids. Established ratios of glucocorticoid metabolites were used to estimate 11ß-HSD1/2 and 5α- and 5ß-reductase activities. RESULTS: While 11ß-HSD1 activity was similar in all groups, 11ß-HSD2 activity was significantly lower in subjects with HNF1A-MODY and T2DM than in healthy controls. The ratio of 5ß- to 5α-metabolites of cortisol was higher in subjects with HNF1A-MODY than in T2DM and healthy controls, probably due to increased activity of the 5ß-reductase (AKR1D1) in HNF1A-MODY. CONCLUSIONS: This is the first report of steroid metabolites in HNF1A-MODY. We have identified distinct differences in steroid metabolism pathways in subjects with HNF1A-MODY that have the potential to alter steroid hormone availability. Further studies are required to explore whether these changes link to phenotype.

Genetics of Monogenic Diabetes: Present Clinical Challenges.

PURPOSE OF REVIEW: Monogenic forms of diabetes have specific treatments that differ from the standard care provided for type 1 and type 2 diabetes, making the appropriate diagnosis essential. In this review, we discuss current clinical challenges that remain, including improving case-finding strategies, particularly those that have transethnic applicability, and understanding the interpretation of genetic variants as pathogenic, with clinically meaningful impacts. RECENT FINDINGS: Biomarker approaches to the stratification for genetic testing now appear to be most effective in identifying cases of monogenic diabetes, and use of genetic risk scores may also prove useful. However, applicability in all ethnic groups is lacking. Challenges remain in the classification of genes as diabetes-causing and the interpretation of genetic variants at the clinical interface. Since the discovery that genetic defects can cause neonatal or young-onset diabetes, multiple causal genes have been identified and there have been many advances in strategies to detect genetic forms of diabetes and their treatments. Approaches learnt from monogenic diabetes are now being translated to polygenic diabetes.

Clustering of cardio-metabolic risk factors in parents of adolescents with type 1 diabetes and microalbuminuria.

OBJECTIVE: To evaluate the association between a clustering of cardio-metabolic risk factors in parents and the development of microalbuminuria (MA) in their offspring with childhood-onset type 1 diabetes (T1D). METHODS: The study population comprised 53 parents (mean age [±SD]: 56.7±6.2 years) of 35 T1D young people with MA (MA+) and 86 parents (age: 56.1±6.3 years) of 50 matched offspring with normoalbuminuria (MA-), who underwent clinical, biochemical and cardiovascular imaging assessments. The primary study endpoint was the difference between parents from the MA+ and MA- groups in a cardio-metabolic risk score, calculated as the average value of the standardized measures (z-scores) for waist circumference, blood pressure, fasting glucose, insulin, HDL-cholesterol and triglycerides levels. Cardiovascular parameters, including carotid intima-media thickness (cIMT), flow-mediated dilatation (FMD) and pulse wave velocity (PWV), were also assessed. A DXA scan was performed to assess body composition. RESULTS: The cardio-metabolic risk score was significantly higher in parents of MA+ compared to parents of MA- offspring (mean [95% CI]: 1.066[0.076; 2.056] vs -0.268[-0.997; 0.460], P = .03). Parents of MA+ offspring had slightly higher values of waist circumference, lipids, insulin and blood pressure, although only diastolic blood pressure was statistically different between the 2 groups (P = .0085). FMD, cIMT, PWV (all P > .3), and DXA parameters (all P > .2) were not significantly different between the 2 groups. CONCLUSIONS: Parents of young offspring with childhood-onset T1D and MA showed an abnormal metabolic profile, reflected by a calculated risk score. The finding supports the role of a familial predisposition to risk of developing diabetic nephropathy.

The Development of Standard Operating Procedures for Social Mobilization and Community Engagement in Sierra Leone During the West Africa Ebola Outbreak of 2014-2015.

This article describes the development of standard operating procedures (SOPs) for social mobilization and community engagement (SM/CE) in Sierra Leone during the Ebola outbreak of 2014-2015. It aims to (a) explain the rationale for a standardized approach, (b) describe the methodology used to develop the resulting SOPs, and (c) discuss the implications of the SOPs for future outbreak responses. Mixed methodologies were applied, including analysis of data on Ebola-related knowledge, attitudes, and practices; consultation through a national forum; and a series of workshops with more than 250 participants active in SM/CE in seven districts with recent confirmed cases. Specific challenges, best practices, and operational models were identified in relation to (a) the quality of SM/CE approaches; (b) coordination and operational structures; and (c) integration with Ebola services, including case management, burials, quarantine, and surveillance. This information was synthesized and codified into the SOPs, which include principles, roles, and actions for partners engaging in SM/CE as part of the Ebola response. This experience points to the need for a set of global principles and standards for meaningful SM/CE that can be rapidly adapted as a high-priority response component at the outset of future health and humanitarian crises.

Glucokinase MODY and implications for treatment goals of common forms of diabetes.

Treatment goals in diabetes concentrate on reducing the risk of vascular complications, largely through setting targets for glycated haemoglobin (HbA1c). These targets are based on epidemiological studies of complication development, but so far have not adequately addressed the adverse effects associated with lowering HbA1c towards the normal range. Glucokinase (GCK) mutations cause a monogenic form of hyperglycaemia (GCK-MODY) characterised by fasting hyperglycaemia with low postprandial glucose excursions and a marginally elevated HbA1c. Minimal levels of vascular complications (comparable with nondiabetic individuals) are observed in GCK-MODY, leading to the hypothesis that GCK-MODY may represent a useful paradigm for assessing treatment goals in all forms of diabetes. In this review, we discuss the evidence behind this concept, suggest ways of translating this hypothesis into clinical practice and address some of the caveats of such an approach.

Transdermal Blood Sampling for C-Peptide Is a Minimally Invasive, Reliable Alternative to Venous Sampling in Children and Adults With Type 1 Diabetes.

OBJECTIVE: C-peptide and islet autoantibodies are key type 1 diabetes biomarkers, typically requiring venous sampling, which limits their utility. We assessed transdermal capillary blood (TCB) collection as a practical alternative. RESEARCH DESIGN AND METHODS: Ninety-one individuals (71 with type 1 diabetes, 20 control; individuals with type 1 diabetes: aged median 14.8 years [interquartile range (IQR) 9.1-17.1], diabetes duration 4.0 years [1.5-7.7]; control individuals: 42.2 years [38.0-52.1]) underwent contemporaneous venous and TCB sampling for measurement of plasma C-peptide. Participants with type 1 diabetes also provided venous serum and plasma, and TCB plasma for measurement of autoantibodies to glutamate decarboxylase, islet antigen-2, and zinc transporter 8. The ability of TCB plasma to detect significant endogenous insulin secretion (venous C-peptide ≥200 pmol/L) was compared along with agreement in levels, using Bland-Altman. Venous serum was compared with venous and TCB plasma for detection of autoantibodies, using established thresholds. Acceptability was assessed by age-appropriate questionnaire. RESULTS: Transdermal sampling took a mean of 2.35 min (SD 1.49). Median sample volume was 50 µL (IQR 40-50) with 3 of 91 (3.3%) failures, and 13 of 88 (14.7%) <35 µL. TCB C-peptide showed good agreement with venous plasma (mean venous ln[C-peptide] - TCB ln[C-peptide] = 0.008, 95% CI [-0.23, 0.29], with 100% [36 of 36] sensitivity/100% [50 of 50] specificity to detect venous C-peptide ≥200 pmol/L). Where venous serum in multiple autoantibody positive TCB plasma agreed in 22 of 32 (sensitivity 69%), comparative specificity was 35 of 36 (97%). TCB was preferred to venous sampling (type 1 diabetes: 63% vs. 7%; 30% undecided). CONCLUSIONS: Transdermal capillary testing for C-peptide is a sensitive, specific, and acceptable alternative to venous sampling; TCB sampling for islet autoantibodies needs further assessment.

PAX6 mutation in association with ptosis, cataract, iris hypoplasia, corneal opacification and diabetes: a new variant of familial aniridia?

BACKGROUND: We report a family with ptosis, cataract, iris hypoplasia and gradual corneal opacification occurring in association with a PAX6 mutation. DESIGN: Case-series. PARTICIPANTS: Fourteen family members - 8 affected, 6 unaffected controls. METHODS: All participants underwent ophthalmological assessment, including best-corrected visual acuity, slit-lamp-examination, pachymetry, endothelial cell-count, tonometry and dilated fundoscopy. All subjects underwent anthropometry and assessment of glycaemic status. Genetic analysis of the PAX6 gene was performed. MAIN OUTCOME MEASURES: Presence of ptosis, corneal, iris and lenticular changes, gycaemic and PAX6 status. RESULTS: All eight affected subjects had ptosis with reduced levator function, anterior polar cataracts, and corneal changes of variable severity - two patients had undergone penetrating keratoplasties, with graft histology revealing conjunctival cells on the cornea and severe fibroinflammatory change. Five patients had iris hypoplasia. One patient had aphakic glaucoma and another had hypoplastic optic discs. Four of the six controls had no ocular features of this syndrome, and two had isolated mild ptosis. There was no difference in height or body mass index between cases and family controls (p > 0.05), but Haemoglobin A1c was greater in the cases (median [interquartile range] 5.6(0.8) vs 5.1(0.3), p = 0.028). Genetic analysis confirmed a pathogenic PAX6 mutation in exon 12 (c1439delC) in all eight patients, but none of the controls. CONCLUSION: This is the first report of this particular constellation of ocular signs occurring in association with a PAX6 mutation. There was no association with anthropometric features, but affected subjects had worse glycaemia than controls, which may be related to the known role of PAX6 in development of the pancreas.

Monogenic diabetes: old and new approaches to diagnosis.

Up to 5% of young adults diagnosed with diabetes have a monogenic aetiology, the most common of which is maturity-onset diabetes of the young (MODY). A definitive molecular diagnosis is important, as this affects treatment, prognosis and family screening. Currently, however, rates of diagnosis are low due to a combination of lack of awareness of the benefits of making the diagnosis and the challenges of differentiating patients with MODY from those with common forms of diabetes. This article aims to introduce general physicians to the characteristics of monogenic diabetes and the clinical features that can be used to diagnose patients. Recently, genomewide association studies have resulted in the identification of C-reactive protein and glycan profile as specific biomarkers for the most common MODY subtype due to HNF1A mutations, and the potential translation of these findings are discussed.

The use of precision diagnostics for monogenic diabetes: a systematic review and expert opinion.

BACKGROUND: Monogenic diabetes presents opportunities for precision medicine but is underdiagnosed. This review systematically assessed the evidence for (1) clinical criteria and (2) methods for genetic testing for monogenic diabetes, summarized resources for (3) considering a gene or (4) variant as causal for monogenic diabetes, provided expert recommendations for (5) reporting of results; and reviewed (6) next steps after monogenic diabetes diagnosis and (7) challenges in precision medicine field. METHODS: Pubmed and Embase databases were searched (1990-2022) using inclusion/exclusion criteria for studies that sequenced one or more monogenic diabetes genes in at least 100 probands (Question 1), evaluated a non-obsolete genetic testing method to diagnose monogenic diabetes (Question 2). The risk of bias was assessed using the revised QUADAS-2 tool. Existing guidelines were summarized for questions 3-5, and review of studies for questions 6-7, supplemented by expert recommendations. Results were summarized in tables and informed recommendations for clinical practice. RESULTS: There are 100, 32, 36, and 14 studies included for questions 1, 2, 6, and 7 respectively. On this basis, four recommendations for who to test and five on how to test for monogenic diabetes are provided. Existing guidelines for variant curation and gene-disease validity curation are summarized. Reporting by gene names is recommended as an alternative to the term MODY. Key steps after making a genetic diagnosis and major gaps in our current knowledge are highlighted. CONCLUSIONS: We provide a synthesis of current evidence and expert opinion on how to use precision diagnostics to identify individuals with monogenic diabetes.

Some diabetes types, called monogenic diabetes, are caused by changes in a single gene. It is important to know who has this kind of diabetes because treatment can differ from that of other types of diabetes. Some treatments also work better than others for specific types, and some people can for example change from insulin injections to tablets. In addition, relatives can be offered a test to see if they are at risk. Genetic testing is needed to diagnose monogenic diabetes but is expensive, so it's not possible to test every person with diabetes for it. We evaluated published research on who should be tested and what test to use. Based on this, we provide recommendations for doctors and health care providers on how to implement genetic testing for monogenic diabetes.

A Systematic Review of the use of Precision Diagnostics in Monogenic Diabetes.

Monogenic forms of diabetes present opportunities for precision medicine as identification of the underlying genetic cause has implications for treatment and prognosis. However, genetic testing remains inconsistent across countries and health providers, often resulting in both missed diagnosis and misclassification of diabetes type. One of the barriers to deploying genetic testing is uncertainty over whom to test as the clinical features for monogenic diabetes overlap with those for both type 1 and type 2 diabetes. In this review, we perform a systematic evaluation of the evidence for the clinical and biochemical criteria used to guide selection of individuals with diabetes for genetic testing and review the evidence for the optimal methods for variant detection in genes involved in monogenic diabetes. In parallel we revisit the current clinical guidelines for genetic testing for monogenic diabetes and provide expert opinion on the interpretation and reporting of genetic tests. We provide a series of recommendations for the field informed by our systematic review, synthesizing evidence, and expert opinion. Finally, we identify major challenges for the field and highlight areas for future research and investment to support wider implementation of precision diagnostics for monogenic diabetes.

Genetic evidence that raised sex hormone binding globulin (SHBG) levels reduce the risk of type 2 diabetes.

Epidemiological studies consistently show that circulating sex hormone binding globulin (SHBG) levels are lower in type 2 diabetes patients than non-diabetic individuals, but the causal nature of this association is controversial. Genetic studies can help dissect causal directions of epidemiological associations because genotypes are much less likely to be confounded, biased or influenced by disease processes. Using this Mendelian randomization principle, we selected a common single nucleotide polymorphism (SNP) near the SHBG gene, rs1799941, that is strongly associated with SHBG levels. We used data from this SNP, or closely correlated SNPs, in 27 657 type 2 diabetes patients and 58 481 controls from 15 studies. We then used data from additional studies to estimate the difference in SHBG levels between type 2 diabetes patients and controls. The SHBG SNP rs1799941 was associated with type 2 diabetes [odds ratio (OR) 0.94, 95% CI: 0.91, 0.97; P = 2 x 10(-5)], with the SHBG raising allele associated with reduced risk of type 2 diabetes. This effect was very similar to that expected (OR 0.92, 95% CI: 0.88, 0.96), given the SHBG-SNP versus SHBG levels association (SHBG levels are 0.2 standard deviations higher per copy of the A allele) and the SHBG levels versus type 2 diabetes association (SHBG levels are 0.23 standard deviations lower in type 2 diabetic patients compared to controls). Results were very similar in men and women. There was no evidence that this variant is associated with diabetes-related intermediate traits, including several measures of insulin secretion and resistance. Our results, together with those from another recent genetic study, strengthen evidence that SHBG and sex hormones are involved in the aetiology of type 2 diabetes.

Maturity onset diabetes of the young: clinical characteristics, diagnosis and management.

Maturity Onset Diabetes of the Young (MODY), represents a rare cause of diabetes (1% of all cases), commonly misdiagnosed as Type 1 Diabetes (T1D) or Type 2 Diabetes (T2D). Clinical characteristics of MODY include age of onset before 45 years, absence of beta-cell autoimmunity and features of metabolic syndrome, sustained endogenous insulin production and strong family history. Common reasons for misdiagnosis are limitations in physicians' awareness and restrictions in performing genetic testing. In an attempt to improve diagnosis rates, recent research efforts have focused on the discovery of non-genetic biomarkers for prioritising individuals for genetic testing, with some promising progress (identification of high-sensitivity CRP, plasma glycan profile as HNF1A-MODY). The information provided is relevant to physicians dealing with young adults but details on pediatric populations are also included. Raising awareness about MODY and making the diagnosis more accessible will improve prognostication and management of these patients and their relatives.

Fucosylated AGP glycopeptides as biomarkers of HNF1A-Maturity onset diabetes of the young.

AIMS: We previously demonstrated that antennary fucosylated N-glycans on plasma proteins are regulated by HNF1A and can identify cases of Maturity-Onset Diabetes of the Young caused by HNF1A variants (HNF1A-MODY). Based on literature data, we further postulated that N-glycans with best diagnostic value mostly originate from alpha-1-acid glycoprotein (AGP). In this study we analyzed fucosylation of AGP in subjects with HNF1A-MODY and other types of diabetes aiming to evaluate its diagnostic potential. METHODS: A recently developed LC-MS method for AGP N-glycopeptide analysis was utilized in two independent cohorts: a) 466 subjects with different diabetes subtypes to test the fucosylation differences, b) 98 selected individuals to test the discriminative potential for pathogenic HNF1A variants. RESULTS: Our results showed significant reduction in AGP fucosylation associated to HNF1A-MODY when compared to other diabetes subtypes. Additionally, ROC curve analysis confirmed significant discriminatory potential of individual fucosylated AGP glycopeptides, where the best performing glycopeptide had an AUC of 0.94 (95% CI 0.90-0.99). CONCLUSIONS: A glycopeptide based diagnostic tool would be beneficial for patient stratification by providing information about the functionality of HNF1A. It could assist the interpretation of DNA sequencing results and be a useful addition to the differential diagnostic process.

Genetics of type 2 diabetes.

Identification and characterization of genetic variants that either cause or predispose to diabetes are a major focus of biomedical research. As of early 2007, the molecular basis of most forms of monogenic diabetes resulting from beta-cell dysfunction is known and, in particular, there has been recent success in delineating the genetic aetiology of neonatal diabetes. Finding genes predisposing to more common, multifactorial forms of type 2 diabetes represents a far greater challenge, and only a handful of robust, well-replicated examples have been established. Nevertheless, 2006 heralded identification of the most important type 2 diabetes susceptibility gene known so far, TCF7L2, and in 2007 large-scale genome-wide association studies are destined to provide novel insights into the genetic architecture and biology of type 2 diabetes.

The correlation between breath acetone and blood betahydroxybutyrate in individuals with type 1 diabetes.

Ketone testing is an important element of the self-management of illness in type 1 diabetes. The aim of the present study was to see if a breath test for acetone could be used to predict quantitatively the levels of the ketone betahydroxybutyrate in the blood of those with type 1 diabetes, and thus be used as an alternative to capillary testing for ketones. Simultaneous capillary ketones and breath acetone were measured in 72 individuals with type 1 diabetes attending a diabetes clinic and on 9 individuals admitted to hospital with diabetic ketoacidosis. Capillary blood measurements ranged from 0.1 mmol l-1 (the lower limit of the ketone monitor) to over 7 mmol l-1, with breath acetone varying between 0.25 and 474 parts per million by volume. The two variables were found to be correlated and allowed modelling to be carried out which separated breath acetone levels into three categories corresponding to normal, elevated and 'at risk' levels of blood ketones. The results on this limited set of participants suggest that a breath acetone test could be a simple, non-invasive substitute for capillary ketone measurement in type 1 diabetes.

Logistic regression has similar performance to optimised machine learning algorithms in a clinical setting: application to the discrimination between type 1 and type 2 diabetes in young adults.

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.

Community engagement in outbreak response: lessons from the 2014-2016 Ebola outbreak in Sierra Leone.

Documentation of structured community engagement initiatives and real-time monitoring of community engagement activities during large-scale epidemics is limited. To inform such initiatives, this paper analyses the Community Led Ebola Action (CLEA) approach implemented through the Social Mobilization Action Consortium (SMAC) during the 2014-2016 Ebola epidemic in Sierra Leone. The SMAC initiative consisted of a network of 2466 community mobilisers, >6000 religious leaders and 42 local radio stations across all 14 districts of Sierra Leone. Community mobilisers were active in nearly 70% of all communities across the country using the CLEA approach to facilitate community analysis, trigger collective action planning and maintain community action plans over time. CLEA was complemented by interactive radio programming and intensified religious leader engagement.Community mobilisers trained in the CLEA approach used participatory methods, comprised of an initial community 'triggering' event, action plan development and weekly follow-ups to monitor progress on identified action items. Mobilisers collected operational and behavioural data on a weekly basis as part of CLEA. We conducted a retrospective analysis of >50 000 weekly reports from approximately 12 000 communities from December 2014 to September 2015. The data showed that 100% of the communities that were engaged had one or more action plans in place. Out of the 63 110 cumulative action points monitored by community mobilisers, 92% were marked as 'in-progress' (85%) or 'achieved' (7%) within 9 months. A qualitative examination of action points revealed that the in-progress status was indicative of the long-term sustainability of most action points (eg, continuous monitoring of visitors into the community) versus one-off action items that were marked as achieved (eg, initial installation of handwashing station). Analysis of behavioural outcomes of the intervention indicate an increase over time in the fraction of reported safe burials and fraction of reported cases referred for medical care within 24 hours of symptom onset in the communities that were engaged.Through CLEA, we have demonstrated how large-scale, coordinated community engagement interventions can be achieved and monitored in real-time during future Ebola epidemics and other similar epidemics. The SMAC initiative provides a practical model for the design, implementation and monitoring of community engagement, integration and coordination of community engagement interventions with other health emergency response pillars, and adaptive strategies for large-scale community-based operational data collection.