The impact of major depressive disorder on glycaemic control in type 2 diabetes: a longitudinal cohort study using UK Biobank primary care records.

BACKGROUND: This study evaluates longitudinal associations between glycaemic control, measured by mean and within-patient variability of glycated haemaglobin (HbA1c) levels, and major depressive disorder (MDD) in individuals with type 2 diabetes (T2D), focusing on the timings of these diagnoses. METHODS: In UK Biobank, T2D was defined using self-report and linked health outcome data, then validated using polygenic scores. Repeated HbA1c measurements (mmol/mol) over the 10 years following T2D diagnosis were outcomes in mixed effects models, with disease duration included using restricted cubic splines. Four MDD exposures were considered: MDD diagnosis prior to T2D diagnosis (pre-T2D MDD), time between pre-T2D MDD diagnosis and T2D, new MDD diagnosis during follow-up (post-T2D MDD) and time since post-T2D MDD diagnosis. Models with and without covariate adjustment were considered. RESULTS: T2D diagnostic criteria were robustly associated with T2D polygenic scores. In 11,837 T2D cases (6.9 years median follow-up), pre-T2D MDD was associated with a 0.92 increase in HbA1c (95% CI: [0.00, 1.84]), but earlier pre-T2D MDD diagnosis correlated with lower HbA1c. These pre-T2D MDD effects became non-significant after covariate adjustment. Post-T2D MDD individuals demonstrated increasing HbA1c with years since MDD diagnosis ( ß = 0.51 , 95% CI: [0.17, 0.86]). Retrospectively, across study follow-up, within-patient variability in HbA1c was 1.16 (95% CI: 1.13-1.19) times higher in post-T2D MDD individuals. CONCLUSIONS: The timing of MDD diagnosis is important for understanding glycaemic control in T2D. Poorer control was observed in MDD diagnosed post-T2D, highlighting the importance of depression screening in T2D, and closer monitoring for individuals who develop MDD after T2D.

Exploring polygenic-environment and residual-environment interactions for depressive symptoms within the UK Biobank.

Substantial advances have been made in identifying genetic contributions to depression, but little is known about how the effect of genes can be modulated by the environment, creating a gene-environment interaction. Using multivariate reaction norm models (MRNMs) within the UK Biobank (N = 61294-91644), we investigate whether the polygenic and residual variance components of depressive symptoms are modulated by 17 a priori selected covariate traits-12 environmental variables and 5 biomarkers. MRNMs, a mixed-effects modelling approach, provide unbiased polygenic-covariate interaction estimates for a quantitative trait by controlling for outcome-covariate correlations and residual-covariate interactions. A continuous depressive symptom variable was the outcome in 17 MRNMs-one for each covariate trait. Each MRNM had a fixed-effects model (fixed effects included the covariate trait, demographic variables, and principal components) and a random effects model (where polygenic-covariate and residual-covariate interactions are modelled). Of the 17 selected covariates, 11 significantly modulate deviations in depressive symptoms through the modelled interactions, but no single interaction explains a large proportion of phenotypic variation. Results are dominated by residual-covariate interactions, suggesting that covariate traits (including neuroticism, childhood trauma, and BMI) typically interact with unmodelled variables, rather than a genome-wide polygenic component, to influence depressive symptoms. Only average sleep duration has a polygenic-covariate interaction explaining a demonstrably nonzero proportion of the variability in depressive symptoms. This effect is small, accounting for only 1.22% (95% confidence interval: [0.54, 1.89]) of variation. The presence of an interaction highlights a specific focus for intervention, but the negative results here indicate a limited contribution from polygenic-environment interactions.

A tool for translating polygenic scores onto the absolute scale using summary statistics.

There is growing interest in the clinical application of polygenic scores as their predictive utility increases for a range of health-related phenotypes. However, providing polygenic score predictions on the absolute scale is an important step for their safe interpretation. We have developed a method to convert polygenic scores to the absolute scale for binary and normally distributed phenotypes. This method uses summary statistics, requiring only the area-under-the-ROC curve (AUC) or variance explained (R2) by the polygenic score, and the prevalence of binary phenotypes, or mean and standard deviation of normally distributed phenotypes. Polygenic scores are converted using normal distribution theory. We also evaluate methods for estimating polygenic score AUC/R2 from genome-wide association study (GWAS) summary statistics alone. We validate the absolute risk conversion and AUC/R2 estimation using data for eight binary and three continuous phenotypes in the UK Biobank sample. When the AUC/R2 of the polygenic score is known, the observed and estimated absolute values were highly concordant. Estimates of AUC/R2 from the lassosum pseudovalidation method were most similar to the observed AUC/R2 values, though estimated values deviated substantially from the observed for autoimmune disorders. This study enables accurate interpretation of polygenic scores using only summary statistics, providing a useful tool for educational and clinical purposes. Furthermore, we have created interactive webtools implementing the conversion to the absolute ( https://opain.github.io/GenoPred/PRS_to_Abs_tool.html ). Several further barriers must be addressed before clinical implementation of polygenic scores, such as ensuring target individuals are well represented by the GWAS sample.

Transforming Summary Statistics from Logistic Regression to the Liability Scale: Application to Genetic and Environmental Risk Scores.

OBJECTIVE: Stratified medicine requires models of disease risk incorporating genetic and environmental factors. These may combine estimates from different studies, and the models must be easily updatable when new estimates become available. The logit scale is often used in genetic and environmental association studies; however, the liability scale is used for polygenic risk scores and measures of heritability, but combining parameters across studies requires a common scale for the estimates. METHODS: We present equations to approximate the relationship between univariate effect size estimates on the logit scale and the liability scale, allowing model parameters to be translated between scales. RESULTS: These equations are used to build a risk score on the liability scale, using effect size estimates originally estimated on the logit scale. Such a score can then be used in a joint effects model to estimate the risk of disease, and this is demonstrated for schizophrenia using a polygenic risk score and environmental risk factors. CONCLUSION: This straightforward method allows the conversion of model parameters between the logit and liability scales and may be a key tool to integrate risk estimates into a comprehensive risk model, particularly for joint models with environmental and genetic risk factors.

Relative risk regression: reliable and flexible methods for log-binomial models.

Relative risks (RRs) are generally considered preferable to odds ratios in prospective studies. However, unlike logistic regression for odds ratios, the standard log-binomial model for RR regression does not respect the natural parameter constraints and is therefore often subject to numerical instability. In this paper, we develop a reliable and flexible method for fitting log-binomial models. We use an Expectation-Maximization (EM) algorithm where the multiplicative event probability is viewed as the joint probability for a collection of latent binary outcomes. This gives a simple iterative scheme that provides stable convergence to the maximum likelihood estimate. In addition to reliability, the method offers some flexible generalizations, including models with unspecified isotonic regression functions. We examine the method's performance using simulations and data analyses of the age-specific RR of mortality following heart attack. These analyses demonstrate the potential for numerical instability in RR regression and show how this can be overcome using the proposed approach. Source code to implement the method in R is provided as supplementary material available at Biostatistics online.