The risk of venous thromboembolism in atopic dermatitis: a matched cohort analysis in UK primary care.

BACKGROUND: Atopic dermatitis (AD) is a common chronic inflammatory skin condition. While other chronic inflammatory conditions are associated with increased risk of venous thromboembolism (VTE), associations between AD and VTE have not been established. OBJECTIVES: We examined whether AD is associated with an increased risk of VTE in a population-based study. METHODS: Electronic health records were extracted from UK general practices contributing to the Optimum Patient Care Research Database (1 January 2010 to 1 January 2020). All adults with AD were identified (n = 150 975) and age- and sex-matched with unaffected controls (n = 603 770). The risk of VTE, consisting of pulmonary embolism (PE) or deep-vein thrombosis (DVT), was compared in people with AD vs. controls using Cox proportional hazard models. PE and DVT were examined separately as secondary outcomes. RESULTS: We identified 150 975 adults with active AD and matched them with 603 770 unaffected controls. During the study, 2576 of those with active AD and 7563 of the matched controls developed VTE. Individuals with AD had a higher risk of VTE than controls [adjusted hazard ratio (aHR) 1.17, 95% confidence interval (CI) 1.12-1.22]. When assessing VTE components, AD was associated with a higher risk of DVT (aHR 1.30, 95% CI 1.23-1.37) but not PE (aHR 0.94, 95% CI 0.87-1.02). The VTE risk was greater in older people with AD (≥ 65 years: aHR 1.22, 95% CI 1.15-1.29; 45-65 years: aHR 1.15, 95% CI 1.05-1.26; < 45 years: aHR 1.07, 95% CI 0.97-1.19) and those with obesity [body mass index (BMI) ≥ 30: aHR 1.25, 95% CI 1.12-1.39; BMI < 30: aHR 1.08, 95% CI 1.01-1.15). Risk was broadly consistent across mild, moderate or severe AD. CONCLUSIONS: AD is associated with a small increase in risk of VTE and DVT, with no increase in risk of PE. The magnitude of this risk increase is modest in younger people, and those without obesity.

Data-Resource Profile: United Kingdom Optimum Patient Care Research Database.

Introduction: Electronic medical records (EMRs) maintained in primary care in the UK and collected and stored in EMR databases offer a world-leading resource for observational clinical research. We aimed to profile one such database: the Optimum Patient Care Research Database (OPCRD). Methods and Participants: The OPCRD, incepted in 2010, is a growing primary care EMR database collecting data from 992 general practices within the UK. It covers over 16.6 million patients across all four countries within the UK, and is broadly representative of the UK population in terms of age, sex, ethnicity and socio-economic status. Patients have a mean duration of 11.7 years' follow-up (SD 17.50), with a majority having key summary data from birth to last data entry. Data for the OPCRD are collected incrementally monthly and extracted from all of the major clinical software systems used within the UK and across all four coding systems (Read version 2, Read CTV3, SNOMED DM+D and SNOMED CT codes). Via quality-improvement programmes provided to GP surgeries, the OPCRD also includes patient-reported outcomes from a range of disease-specific validated questionnaires, with over 66,000 patient responses on asthma, COPD, and COVID-19. Further, bespoke data collection is possible by working with GPs to collect new research via patient-reported questionnaires. Findings to Date: The OPCRD has contributed to over 96 peer-reviewed research publications since its inception encompassing a broad range of medical conditions, including COVID-19. Conclusion: The OPCRD represents a unique resource with great potential to support epidemiological research, from retrospective observational studies through to embedded cluster-randomised trials. Advantages of the OPCRD over other EMR databases are its large size, UK-wide geographical coverage, the availability of up-to-date patient data from all major GP software systems, and the unique collection of patient-reported information on respiratory health.

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.

Development and validation of multivariable clinical diagnostic models to identify type 1 diabetes requiring rapid insulin therapy in adults aged 18-50 years.

OBJECTIVE: To develop and validate multivariable clinical diagnostic models to assist distinguishing between type 1 and type 2 diabetes in adults aged 18-50. DESIGN: Multivariable logistic regression analysis was used to develop classification models integrating five pre-specified predictor variables, including clinical features (age of diagnosis, body mass index) and clinical biomarkers (GADA and Islet Antigen 2 islet autoantibodies, Type 1 Diabetes Genetic Risk Score), to identify type 1 diabetes with rapid insulin requirement using data from existing cohorts. SETTING: UK cohorts recruited from primary and secondary care. PARTICIPANTS: 1352 (model development) and 582 (external validation) participants diagnosed with diabetes between the age of 18 and 50 years of white European origin. MAIN OUTCOME MEASURES: Type 1 diabetes was defined by rapid insulin requirement (within 3 years of diagnosis) and severe endogenous insulin deficiency (C-peptide <200 pmol/L). Type 2 diabetes was defined by either a lack of rapid insulin requirement or, where insulin treated within 3 years, retained endogenous insulin secretion (C-peptide >600 pmol/L at ≥5 years diabetes duration). Model performance was assessed using area under the receiver operating characteristic curve (ROC AUC), and internal and external validation. RESULTS: Type 1 diabetes was present in 13% of participants in the development cohort. All five predictor variables were discriminative and independent predictors of type 1 diabetes (p<0.001 for all) with individual ROC AUC ranging from 0.82 to 0.85. Model performance was high: ROC AUC range 0.90 (95% CI 0.88 to 0.93) (clinical features only) to 0.97 (95% CI 0.96 to 0.98) (all predictors) with low prediction error. Results were consistent in external validation (clinical features and GADA ROC AUC 0.93 (0.90 to 0.96)). CONCLUSIONS: Clinical diagnostic models integrating clinical features with biomarkers have high accuracy for identifying type 1 diabetes with rapid insulin requirement, and could assist clinicians and researchers in accurately identifying patients with type 1 diabetes.

Type 1 diabetes defined by severe insulin deficiency occurs after 30 years of age and is commonly treated as type 2 diabetes.

AIMS/HYPOTHESIS: Late-onset type 1 diabetes can be difficult to identify. Measurement of endogenous insulin secretion using C-peptide provides a gold standard classification of diabetes type in longstanding diabetes that closely relates to treatment requirements. We aimed to determine the prevalence and characteristics of type 1 diabetes defined by severe endogenous insulin deficiency after age 30 and assess whether these individuals are identified and managed as having type 1 diabetes in clinical practice. METHODS: We assessed the characteristics of type 1 diabetes defined by rapid insulin requirement (within 3 years of diagnosis) and severe endogenous insulin deficiency (non-fasting C-peptide <200 pmol/l) in 583 participants with insulin-treated diabetes, diagnosed after age 30, from the Diabetes Alliance for Research in England (DARE) population cohort. We compared characteristics with participants with retained endogenous insulin secretion (>600 pmol/l) and 220 participants with severe insulin deficiency who were diagnosed under age 30. RESULTS: Twenty-one per cent of participants with insulin-treated diabetes who were diagnosed after age 30 met the study criteria for type 1 diabetes. Of these participants, 38% did not receive insulin at diagnosis, of whom 47% self-reported type 2 diabetes. Rapid insulin requirement was highly predictive of severe endogenous insulin deficiency: 85% required insulin within 1 year of diagnosis, and 47% of all those initially treated without insulin who progressed to insulin treatment within 3 years of diagnosis had severe endogenous insulin deficiency. Participants with late-onset type 1 diabetes defined by development of severe insulin deficiency had similar clinical characteristics to those with young-onset type 1 diabetes. However, those with later onset type 1 diabetes had a modestly lower type 1 diabetes genetic risk score (0.268 vs 0.279; p < 0.001 [expected type 2 diabetes population median, 0.231]), a higher islet autoantibody prevalence (GAD-, islet antigen 2 [IA2]- or zinc transporter protein 8 [ZnT8]-positive) of 78% at 13 years vs 62% at 26 years of diabetes duration; (p = 0.02), and were less likely to identify as having type 1 diabetes (79% vs 100%; p < 0.001) vs those with young-onset disease. CONCLUSIONS/INTERPRETATION: Type 1 diabetes diagnosed over 30 years of age, defined by severe insulin deficiency, has similar clinical and biological characteristics to that occurring at younger ages, but is frequently not identified. Clinicians should be aware that patients progressing to insulin within 3 years of diagnosis have a high likelihood of type 1 diabetes, regardless of initial diagnosis.