Mapping of UK Biobank clinical codes: Challenges and possible solutions

Causal associations of central and peripheral risk factors with knee osteoarthritis: a longitudinal and Mendelian Randomisation study using UK Biobank data

ABSTRACT

Our aim was to investigate relative contributions of central and peripheral mechanisms to knee osteoarthritis (OA) diagnosis and their independent causal association with knee OA. We performed longitudinal analysis using data from UK-Biobank participants. Knee OA was defined using International Classification of Diseases manual 10 codes from participants' hospital records. Central mechanisms were proxied using multisite chronic pain (MCP) and peripheral mechanisms using body mass index (BMI). Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated, and proportional risk contribution (PRC) was estimated from receiver-operator-characteristic (ROC) analysis. To estimate the causal effects, we performed 2-sample multivariable Mendelian Randomisation (MR) analysis. We selected genetic instruments from the largest Genome Wide Association Study of BMI (N = 806,834) and MCP (N = 387,649) and estimated the instruments genetic associations with knee OA in the largest available dataset (62,497 cases and 333,557 control subjects). The multivariable MR was performed using modified inverse-variance weighting methods. Of the 203,410 participants, 6% developed knee OA. Both MCP (OR 1.23, 95% CI; 1.21-1.24) and BMI (1.10, 95% CI; 1.10-1.11) were associated with knee OA diagnosis. The PRC was 6.9% (95% CI; 6.7%-7.1%) for MCP and 21.9% (95% CI; 21.4%-22.5%) for BMI; the combined PRC was 38.8% (95% CI; 37.9%-39.8%). Body mass index and MCP had independent causal effects on knee OA (OR 1.76 [95% CI, 1.64-1.88] and 1.83 [95% CI, 1.54-2.16] per unit change, respectively). In conclusion, peripheral risk factors (eg, BMI) contribute more to the development of knee OA than central risk factors (eg, MCP). Peripheral and central factors are independently causal on knee OA.

Association of whole-person eigen-polygenic risk scores with Alzheimer's disease

Late-Onset Alzheimer's Disease (LOAD) is a heterogeneous neurodegenerative disorder with complex etiology and high heritability. Its multifactorial risk profile and large portions of unexplained heritability suggest the involvement of yet unidentified genetic risk factors. Here we describe the "whole person" genetic risk landscape of polygenic risk scores for 2218 traits in 2044 elderly individuals and test if novel eigen-PRSs derived from clustered subnetworks of single-trait PRSs can improve the prediction of LOAD diagnosis, rates of cognitive decline, and canonical LOAD neuropathology. Network analyses revealed distinct clusters of PRSs with clinical and biological interpretability. Novel eigen-PRSs (ePRS) from these clusters significantly improved LOAD-related phenotypes prediction over current state-of-the-art LOAD PRS models. Notably, an ePRS representing clusters of traits related to cholesterol levels was able to improve variance explained in a model of the brain-wide beta-amyloid burden by 1.7% (likelihood ratio test P = 9.02 × 10-7). All associations of ePRS with LOAD phenotypes were eliminated by the removal of APOE-proximal loci. However, our association analysis identified modules characterized by PRSs of high cholesterol and LOAD. We believe this is due to the influence of the APOE region from both PRSs. We found significantly higher mean SNP effects for LOAD in the intersecting APOE region SNPs. Combining genetic risk factors for vascular traits and dementia could improve current single-trait PRS models of LOAD, enhancing the use of PRS in risk stratification. Our results are catalogued for the scientific community, to aid in generating new hypotheses based on our maps of clustered PRSs and associations with LOAD-related phenotypes.

Opaque Ontology: Neuroimaging Classification of ICD-10 Diagnostic Groups in the UK Biobank

Background

1.The use of machine learning to classify diagnostic cases versus controls defined based on diagnostic ontologies such as the ICD-10 from neuroimaging features is now commonplace across a wide range of diagnostic fields. However, transdiagnostic comparisons of such classifications are lacking. Such transdiagnostic comparisons are important to establish the specificity of classification models, set benchmarks, and assess the value of diagnostic ontologies.

Results

2.We investigated case-control classification accuracy in 17 different ICD-10 diagnostic groups from Chapter V (mental and behavioral disorders) and Chapter VI (diseases of the nervous system) using data from the UK Biobank. Classification models were trained using either neuroimaging (structural or functional brain MRI feature sets) or socio-demographic features. Random forest classification models were adopted using rigorous shuffle splits to estimate stability as well as accuracy of case-control classifications. Diagnostic classification accuracies were benchmarked against age classification (oldest versus youngest) from the same feature sets and against additional classifier types (K-nearest neighbors and linear support vector machine). In contrast to age classification accuracy, which was high for all feature sets, few ICD-10 diagnostic groups were classified significantly above chance (namely, demyelinating diseases based on structural neuroimaging features, and depression based on socio-demographic and functional neuroimaging features).

Conclusion

3.These findings highlight challenges with the current disease classification system, leading us to recommend caution with the use of ICD-10 diagnostic groups as target labels in brain-based disease prediction studies.

Development and validation of a lifetime prediction model for incident type 2 diabetes in patients with established cardiovascular disease: the CVD2DM model

AIMS

Identifying patients with established cardiovascular disease (CVD) who are at high risk of type 2 diabetes (T2D) may allow for early interventions, reducing the development of T2D and associated morbidity. The aim of this study was to develop and externally validate the CVD2DM model to estimate the 10-year and lifetime risks of T2D in patients with established CVD.

METHODS AND RESULTS

Sex-specific, competing risk-adjusted Cox proportional hazard models were derived in 19 281 participants with established CVD and without diabetes at baseline from the UK Biobank. The core model's pre-specified predictors were age, current smoking, family history of diabetes mellitus, body mass index, systolic blood pressure, fasting plasma glucose, and HDL cholesterol. The extended model also included HbA1c. The model was externally validated in 3481 patients from the UCC-SMART study. During a median follow-up of 12.2 years (interquartile interval 11.3-13.1), 1628 participants with established CVD were diagnosed with T2D in the UK Biobank. External validation c-statistics were 0.79 [95% confidence interval (CI) 0.76-0.82] for the core model and 0.81 (95% CI 0.78-0.84) for the extended model. Calibration plots showed agreement between predicted and observed 10-year risk of T2D.

CONCLUSION

The 10-year and lifetime risks of T2D can be estimated with the CVD2DM model in patients with established CVD, using readily available clinical predictors. The model would benefit from further validation across diverse ethnic groups to enhance its applicability. Informing patients about their T2D risk could motivate them further to adhere to a healthy lifestyle.

Association between pre-dementia psychiatric diagnoses and all-cause dementia is independent from polygenic dementia risks in the UK Biobank

BACKGROUND

Psychiatric disorders have been associated with higher risk for future dementia. Understanding how pre-dementia psychiatric disorders (PDPD) relate to established dementia genetic risks has implications for dementia prevention.

METHODS

In this retrospective cohort study, we investigated the relationships between polygenic risk scores for Alzheimer's disease (AD PRS), PDPD, alcohol use disorder (AUD), and subsequent dementia in the UK Biobank (UKB) and tested whether the relationships are consistent with different causal models.

FINDINGS

Among 502,408 participants, 9352 had dementia. As expected, AD PRS was associated with greater risk for dementia (odds ratio (OR) 1.62, 95% confidence interval (CI), 1.59-1.65). A total of 94,237 participants had PDPD, of whom 2.6% (n = 2519) developed subsequent dementia, compared to 1.7% (n = 6833) of 407,871 participants without PDPD. Accordingly, PDPD were associated with 73% greater risk of incident dementia (OR 1.73, 1.65-1.83). Among dementia subtypes, the risk increase was 1.5-fold for AD (n = 3365) (OR 1.46, 1.34-1.59) and 2-fold for vascular dementia (VaD, n = 1823) (OR 2.08, 1.87-2.32). Our data indicated that PDPD were neither a dementia prodrome nor a mediator for AD PRS. Shared factors for both PDPD and dementia likely substantially account for the observed association, while a causal role of PDPD in dementia could not be excluded. AUD could be one of the shared causes for PDPD and dementia.

INTERPRETATION

Psychiatric diagnoses were associated with subsequent dementia in UKB participants, and the association is orthogonal to established dementia genetic risks. Investigating shared causes for psychiatric disorders and dementia would shed light on this dementia pathway.

FUNDING

US NIH (K08AG054727).

Association between cancer and dementia risk in the UK Biobank: evidence of diagnostic bias

Epidemiological studies have identified an inverse association between cancer and dementia. Underlying methodological biases have been postulated, yet no studies have systematically investigated the potential for each source of bias within a single dataset. We used the UK Biobank to compare estimates for the cancer-dementia association using different analytical specifications designed to sequentially address multiple sources of bias, including competing risk of death, selective survival, confounding bias, and diagnostic bias. We included 140,959 UK Biobank participants aged ≥ 55 without dementia before enrollment and with linked primary care data. We used cancer registry data to identify cancer cases prevalent before UK Biobank enrollment and incident cancer diagnosed after enrollment. We used Cox models to evaluate associations of prevalent and incident cancer with all-cause dementia, Alzheimer's disease (AD), and vascular dementia. We used time-varying models to evaluate diagnostic bias. Over a median follow-up of 12.3 years, 3,310 dementia cases were diagnosed. All-site incident cancer was positively associated with all-cause dementia incidence (hazard ratio [HR] = 1.14, 95% CI: 1.02-1.29), but prevalent cancer was not (HR = 1.04, 95% CI: 0.92-1.17). Results were similar for vascular dementia. AD was not associated with prevalent or incident cancer. Dementia diagnosis was substantially elevated in the first year after cancer diagnosis (HR = 1.83, 95% CI: 1.42-2.36), after which the association attenuated to null, suggesting diagnostic bias. Following a cancer diagnosis, health care utilization or cognitive consequences of diagnosis or treatment may increase chance of receiving a dementia diagnosis, creating potential diagnostic bias in electronic health records-based studies.