Inferring causation from non-randomized studies of exposure requires that exposure groups can be balanced with respect to prognostic factors for the outcome. Although there is broad agreement in the literature that balance should be checked, there is confusion regarding the appropriate metric. We present a simulation study that compares several balance metrics with respect to the strength of their association with bias in estimation of the effect of a binary exposure on a binary, count, or continuous outcome. The simulations utilize matching on the propensity score with successively decreasing calipers to produce datasets with varying covariate balance. We propose the post-matching C-statistic as a balance metric and found that it had consistently strong associations with estimation bias, even when the propensity score model was misspecified, as long as the propensity score was estimated with sufficient study size. This metric, along with the average standardized difference and the general weighted difference, outperformed all other metrics considered in association with bias, including the unstandardized absolute difference, Kolmogorov-Smirnov and Lévy distances, overlapping coefficient, Mahalanobis balance, and L1 metrics. Of the best-performing metrics, the C-statistic and general weighted difference also have the advantage that they automatically evaluate balance on all covariates simultaneously and can easily incorporate balance on interactions among covariates. Therefore, when combined with the usual practice of comparing individual covariate means and standard deviations across exposure groups, these metrics may provide useful summaries of the observed covariate imbalance.

No abstract

ObjectivesAtrial fibrillation (AF) is the most common cardiac rhythm disorder with a significant health burden. The aim of this study was to characterise patients with recently diagnosed AF and to estimate the rates of comorbidities and outcome events requiring hospitalisation in routine clinical practice.DesignPharmacoepidemiological cohort study using observational data.Methods/settingThis study included 16 513 patients with a first diagnosis of AF between 1 January 2005 and 28 February 2010 (newly diagnosed patients) using data from the UK Clinical Practice Research Datalink (CPRD) linked to Hospital Episode Statistics (HES) and the Office for National Statistics mortality data. Exposure was stratified by vitamin K antagonist (VKA) exposure (non-use, current, recent and past exposure) based on prescriptions and/or international normalised ratio measurements, and followed for outcome events of interest based on diagnosis codes in the databases, that is, vascular outcomes, bleeding events and others. The main focus of the study was on outcome events requiring hospitalisation using the HES data.ResultsThe incidence of vascular outcome hospitalisations (myocardial infarction (MI), stroke or systemic arterial peripheral embolism) was 3.8 (95% CI 3.5 to 4.0)/100 patient-years. The incidence of stroke was 0.9 (0.8 to 1.1) during current VKA exposure, 2.2 (1.6 to 2.9) for recent, 2.4 (1.9 to 2.9) for past and 3.4 (3.1 to 3.7) during non-use. MI incidence was 0.7 (0.6 to 0.9) for current VKA exposure, 0.7 (0.4 to 1.2) for recent, 1.1 (0.8 to 1.5) for past and 1.9 (1.7 to 2.1) during non-use. The incidence of bleeding event hospitalisations was 3.8 (3.4 to 4.2) for current VKA exposure, 4.5 (3.7 to 5.5) for recent, 2.7 (2.2 to 3.3) for past and 2.9 (2.6 to 3.2) during non-use; 38% of intracranial bleeds and 6% of gastrointestinal bleeds were fatal.ConclusionsThis population-based study from recent years provides a comprehensive characterisation of newly diagnosed patients with AF and incidence estimates of common outcomes with a focus on hospitalised events stratified by VKA exposure. This study will help to place future data on new oral anticoagulants into perspective.

Background/Aims: Dialysis patients display an increased mortality which is associated with cardiovascular calcifications. Diabetes mellitus and ethnicity are known factors that affect the extent of cardiovascular calcifications. However, most studies have investigated mixed cohorts with diabetics and/or mixed ethnicity. Methods: Cardiovascular calcifications were assessed in non-diabetic Caucasian haemodialysis patients by the semiquantitative Adragao calcification score (X-ray pelvis and hands) and a novel composite calcification score encompassing the Adragao score as well as calcifications detected by X-ray of the fistula arm, echocardiography of heart valves and carotid ultrasound. Results: Using multivariate analysis, age, male gender, dialysis vintage, lower Kt/V, calcium-phosphate product, smoking and high-sensitivity CRP were independent risk factors for cardiovascular calcifications as assessed by the Adragao or the composite score. Pulse wave velocity was independently related to both calcification scores. Body mass index, cholesterol, triglycerides, iPTH and serum levels of fetuin-A and uncarboxylated matrix Gla protein were not associated with cardiovascular calcifications. Conclusions: In our cohort of non-diabetic Caucasian haemodialysis patients, age, male gender, dialysis vintage, smoking, calcium-phosphate product, high-sensitivity CRP and lower Kt/V were independent risk factors for cardiovascular calcifications. Whether lowering the calcium-phosphate product and increasing dialysis efficiency can reduce cardiovascular calcifications in dialysis patients remains to be determined.