Sharing Medical Big Data While Preserving Patient Confidentiality in Innovative Medicines Initiative: A Summary and Case Report from BigData@Heart.

Sharing individual patient data (IPD) is a simple concept but complex to achieve due to data privacy and data security concerns, underdeveloped guidelines, and legal barriers. Sharing IPD is additionally difficult in big data-driven collaborations such as Bigdata@Heart in the Innovative Medicines Initiative, due to competing interests between diverse consortium members. One project within BigData@Heart, case study 1, needed to pool data from seven heterogeneous data sets: five randomized controlled trials from three different industry partners, and two disease registries. Sharing IPD was not considered feasible due to legal requirements and the sensitive medical nature of these data. In addition, harmonizing the data sets for a federated data analysis was difficult due to capacity constraints and the heterogeneity of the data sets. An alternative option was to share summary statistics through contingency tables. Here it is demonstrated that this method along with anonymization methods to ensure patient anonymity had minimal loss of information. Although sharing IPD should continue to be encouraged and strived for, our approach achieved a good balance between data transparency while protecting patient privacy. It also allowed a successful collaboration between industry and academia.

Environmental and genetic predictors of human cardiovascular ageing.

Cardiovascular ageing is a process that begins early in life and leads to a progressive change in structure and decline in function due to accumulated damage across diverse cell types, tissues and organs contributing to multi-morbidity. Damaging biophysical, metabolic and immunological factors exceed endogenous repair mechanisms resulting in a pro-fibrotic state, cellular senescence and end-organ damage, however the genetic architecture of cardiovascular ageing is not known. Here we use machine learning approaches to quantify cardiovascular age from image-derived traits of vascular function, cardiac motion and myocardial fibrosis, as well as conduction traits from electrocardiograms, in 39,559 participants of UK Biobank. Cardiovascular ageing is found to be significantly associated with common or rare variants in genes regulating sarcomere homeostasis, myocardial immunomodulation, and tissue responses to biophysical stress. Ageing is accelerated by cardiometabolic risk factors and we also identify prescribed medications that are potential modifiers of ageing. Through large-scale modelling of ageing across multiple traits our results reveal insights into the mechanisms driving premature cardiovascular ageing and reveal potential molecular targets to attenuate age-related processes.

Genetic studies of paired metabolomes reveal enzymatic and transport processes at the interface of plasma and urine.

The kidneys operate at the interface of plasma and urine by clearing molecular waste products while retaining valuable solutes. Genetic studies of paired plasma and urine metabolomes may identify underlying processes. We conducted genome-wide studies of 1,916 plasma and urine metabolites and detected 1,299 significant associations. Associations with 40% of implicated metabolites would have been missed by studying plasma alone. We detected urine-specific findings that provide information about metabolite reabsorption in the kidney, such as aquaporin (AQP)-7-mediated glycerol transport, and different metabolomic footprints of kidney-expressed proteins in plasma and urine that are consistent with their localization and function, including the transporters NaDC3 (SLC13A3) and ASBT (SLC10A2). Shared genetic determinants of 7,073 metabolite-disease combinations represent a resource to better understand metabolic diseases and revealed connections of dipeptidase 1 with circulating digestive enzymes and with hypertension. Extending genetic studies of the metabolome beyond plasma yields unique insights into processes at the interface of body compartments.

Prevalence, phenotypic characteristics and prognostic role of apparent treatment resistant hypertension in the German Chronic Kidney Disease (GCKD) study.

Treatment resistant hypertension (TRH) appears of particular relevance in patients with chronic kidney disease (CKD). However, causes and consequences of TRH in CKD patients remain incompletely understood. Therefore, we analyzed the prevalence of apparent TRH (aTRH), and phenotypic characteristics and prognosis associated with aTRH among participants of the German Chronic Kidney Disease (GCKD) study. As insufficient medication adherence has been shown to be a frequent cause of pseudoresistance, we also assessed treatment adherence. Study participants were classified as having aTRH, controlled hypertension and uncontrolled hypertension based on study visit blood pressure and self-reported medication intake. Drug adherence was assessed by comparing self-reported antihypertensive medication with detectable urinary drug metabolites measured by mass spectroscopy. Out of 4901 individuals included in this study, 38% were classified as having aTRH. Male sex, older age, lower estimated glomerular filtration rate (eGFR), higher body mass index (BMI), higher urine albumin-to-creatinine ratio (UACR) and presence of diabetes mellitus were independently associated with higher prevalence of aTRH in a multivariable adjusted regression model. Patients classified as aTRH had higher risk for major adverse cardiovascular events and worsening of kidney disease compared to patients with no aTRH after multivariate adjustment for potential confounders. There was a high agreement between self-reported medication and detectable urinary drug metabolites. In conclusion, in a cohort of Caucasian patients with moderately severe CKD, aTRH was highly prevalent and, in most cases, likely not caused by low medication adherence. Furthermore, aTRH was linked to cardio-renal endpoints, emphasizing the need for improved management.

Deep learning on resting electrocardiogram to identify impaired heart rate recovery.

Background and Objective: Postexercise heart rate recovery (HRR) is an important indicator of cardiac autonomic function and abnormal HRR is associated with adverse outcomes. We hypothesized that deep learning on resting electrocardiogram (ECG) tracings may identify individuals with impaired HRR. Methods: We trained a deep learning model (convolutional neural network) to infer HRR based on resting ECG waveforms (HRRpred) among UK Biobank participants who had undergone exercise testing. We examined the association of HRRpred with incident cardiovascular disease using Cox models, and investigated the genetic architecture of HRRpred in genome-wide association analysis. Results: Among 56,793 individuals (mean age 57 years, 51% women), the HRRpred model was moderately correlated with actual HRR (r = 0.48, 95% confidence interval [CI] 0.47-0.48). Over a median follow-up of 10 years, we observed 2060 incident diabetes mellitus (DM) events, 862 heart failure events, and 2065 deaths. Higher HRRpred was associated with lower risk of DM (hazard ratio [HR] 0.79 per 1 standard deviation change, 95% CI 0.76-0.83), heart failure (HR 0.89, 95% CI 0.83-0.95), and death (HR 0.83, 95% CI 0.79-0.86). After accounting for resting heart rate, the association of HRRpred with incident DM and all-cause mortality were similar. Genetic determinants of HRRpred included known heart rate, cardiac conduction system, cardiomyopathy, and metabolic trait loci. Conclusion: Deep learning-derived estimates of HRR using resting ECG independently associated with future clinical outcomes, including new-onset DM and all-cause mortality. Inferring postexercise heart rate response from a resting ECG may have potential clinical implications and impact on preventive strategies warrants future study.

Cohort design and natural language processing to reduce bias in electronic health records research.

Electronic health record (EHR) datasets are statistically powerful but are subject to ascertainment bias and missingness. Using the Mass General Brigham multi-institutional EHR, we approximated a community-based cohort by sampling patients receiving longitudinal primary care between 2001-2018 (Community Care Cohort Project [C3PO], n = 520,868). We utilized natural language processing (NLP) to recover vital signs from unstructured notes. We assessed the validity of C3PO by deploying established risk models for myocardial infarction/stroke and atrial fibrillation. We then compared C3PO to Convenience Samples including all individuals from the same EHR with complete data, but without a longitudinal primary care requirement. NLP reduced the missingness of vital signs by 31%. NLP-recovered vital signs were highly correlated with values derived from structured fields (Pearson r range 0.95-0.99). Atrial fibrillation and myocardial infarction/stroke incidence were lower and risk models were better calibrated in C3PO as opposed to the Convenience Samples (calibration error range for myocardial infarction/stroke: 0.012-0.030 in C3PO vs. 0.028-0.046 in Convenience Samples; calibration error for atrial fibrillation 0.028 in C3PO vs. 0.036 in Convenience Samples). Sampling patients receiving regular primary care and using NLP to recover missing data may reduce bias and maximize generalizability of EHR research.

Genetic and environmental determinants of diastolic heart function.

Diastole is the sequence of physiological events that occur in the heart during ventricular filling and principally depends on myocardial relaxation and chamber stiffness. Abnormal diastolic function is related to many cardiovascular disease processes and is predictive of health outcomes, but its genetic architecture is largely unknown. Here, we use machine learning cardiac motion analysis to measure diastolic functional traits in 39,559 participants of the UK Biobank and perform a genome-wide association study. We identified 9 significant, independent loci near genes that are associated with maintaining sarcomeric function under biomechanical stress and genes implicated in the development of cardiomyopathy. Age, sex and diabetes were independent predictors of diastolic function and we found a causal relationship between genetically-determined ventricular stiffness and incident heart failure. Our results provide insights into the genetic and environmental factors influencing diastolic function that are relevant for identifying causal relationships and potential tractable targets.

Batch Selection via In Vitro/In Vivo Correlation in Pharmacokinetic Bioequivalence Testing.

Pharmacokinetic differences between manufacturing batches, well established for inhaled drug products, preclude control of patient risk in the customary two-way (single batch) pharmacokinetic bioequivalence crossover design if batches are randomly chosen. European regulators have recommended selecting a "typical" in vitro batch to represent each product in pharmacokinetic bioequivalence testing. We explored the feasibility of this approach to control patient risk (the "false equivalence", or Type I, error rate). The probability of achieving a Test/Reference 90% confidence interval within (0.80, 1.25) for a true (non-equivalent) value of 1.25 was simulated for a two-way crossover design using the median in vitro batch across a range of number of in vitro batches, in vitro/in vivo correlation (IVIVC) quality (correlation coefficient, r, of zero to one), and within-subject between-batch pharmacokinetic variability. Even under extremely optimistic conditions, e.g., r=0.95 and >100 batches per product screened in vitro, patient risk for typical between-batch variability levels remained at least threefold higher than the 5% regulatory expectation for the significance level (the false equivalence error rate) of the pharmacokinetic bioequivalence test. This elevated error rate in bioequivalence decision-making occurs because of incomplete confidence that the true product average has been identified, and, importantly, omission of this uncertainty from the bioequivalence confidence interval.

Self-Reported Medication Use and Urinary Drug Metabolites in the German Chronic Kidney Disease (GCKD) Study.

BACKGROUND: Polypharmacy is common among patients with CKD, but little is known about the urinary excretion of many drugs and their metabolites among patients with CKD. METHODS: To evaluate self-reported medication use in relation to urine drug metabolite levels in a large cohort of patients with CKD, the German Chronic Kidney Disease study, we ascertained self-reported use of 158 substances and 41 medication groups, and coded active ingredients according to the Anatomical Therapeutic Chemical Classification System. We used a nontargeted mass spectrometry-based approach to quantify metabolites in urine; calculated specificity, sensitivity, and accuracy of medication use and corresponding metabolite measurements; and used multivariable regression models to evaluate associations and prescription patterns. RESULTS: Among 4885 participants, there were 108 medication-drug metabolite pairs on the basis of reported medication use and 78 drug metabolites. Accuracy was excellent for measurements of 36 individual substances in which the unchanged drug was measured in urine (median, 98.5%; range, 61.1%-100%). For 66 pairs of substances and their related drug metabolites, median measurement-based specificity and sensitivity were 99.2% (range, 84.0%-100%) and 71.7% (range, 1.2%-100%), respectively. Commonly prescribed medications for hypertension and cardiovascular risk reduction-including angiotensin II receptor blockers, calcium channel blockers, and metoprolol-showed high sensitivity and specificity. Although self-reported use of prescribed analgesics (acetaminophen, ibuprofen) was <3% each, drug metabolite levels indicated higher usage (acetaminophen, 10%-26%; ibuprofen, 10%-18%). CONCLUSIONS: This comprehensive screen of associations between urine drug metabolite levels and self-reported medication use supports the use of pharmacometabolomics to assess medication adherence and prescription patterns in persons with CKD, and indicates under-reported use of medications available over the counter, such as analgesics.

Machine learning for subtype definition and risk prediction in heart failure, acute coronary syndromes and atrial fibrillation: systematic review of validity and clinical utility.

BACKGROUND: Machine learning (ML) is increasingly used in research for subtype definition and risk prediction, particularly in cardiovascular diseases. No existing ML models are routinely used for cardiovascular disease management, and their phase of clinical utility is unknown, partly due to a lack of clear criteria. We evaluated ML for subtype definition and risk prediction in heart failure (HF), acute coronary syndromes (ACS) and atrial fibrillation (AF). METHODS: For ML studies of subtype definition and risk prediction, we conducted a systematic review in HF, ACS and AF, using PubMed, MEDLINE and Web of Science from January 2000 until December 2019. By adapting published criteria for diagnostic and prognostic studies, we developed a seven-domain, ML-specific checklist. RESULTS: Of 5918 studies identified, 97 were included. Across studies for subtype definition (n = 40) and risk prediction (n = 57), there was variation in data source, population size (median 606 and median 6769), clinical setting (outpatient, inpatient, different departments), number of covariates (median 19 and median 48) and ML methods. All studies were single disease, most were North American (n = 61/97) and only 14 studies combined definition and risk prediction. Subtype definition and risk prediction studies respectively had limitations in development (e.g. 15.0% and 78.9% of studies related to patient benefit; 15.0% and 15.8% had low patient selection bias), validation (12.5% and 5.3% externally validated) and impact (32.5% and 91.2% improved outcome prediction; no effectiveness or cost-effectiveness evaluations). CONCLUSIONS: Studies of ML in HF, ACS and AF are limited by number and type of included covariates, ML methods, population size, country, clinical setting and focus on single diseases, not overlap or multimorbidity. Clinical utility and implementation rely on improvements in development, validation and impact, facilitated by simple checklists. We provide clear steps prior to safe implementation of machine learning in clinical practice for cardiovascular diseases and other disease areas.

Rare genetic variants affecting urine metabolite levels link population variation to inborn errors of metabolism.

Metabolite levels in urine may provide insights into genetic mechanisms shaping their related pathways. We therefore investigate the cumulative contribution of rare, exonic genetic variants on urine levels of 1487 metabolites and 53,714 metabolite ratios among 4864 GCKD study participants. Here we report the detection of 128 significant associations involving 30 unique genes, 16 of which are known to underlie inborn errors of metabolism. The 30 genes are strongly enriched for shared expression in liver and kidney (odds ratio = 65, p-FDR = 3e-7), with hepatocytes and proximal tubule cells as driving cell types. Use of UK Biobank whole-exome sequencing data links genes to diseases connected to the identified metabolites. In silico constraint-based modeling of gene knockouts in a virtual whole-body, organ-resolved metabolic human correctly predicts the observed direction of metabolite changes, highlighting the potential of linking population genetics to modeling. Our study implicates candidate variants and genes for inborn errors of metabolism.

Leveraging Human Genetics to Estimate Clinical Risk Reductions Achievable by Inhibiting Factor XI.

Background and Purpose- Coagulation factor XI (FXI) is a novel target for antithrombotic therapy addressed by various therapeutic modalities currently in clinical development. The expected magnitude of thrombotic event reduction mediated by targeting FXI is unclear. Methods- We analyzed the association of 2 common genetic variants, which alter levels of FXI, with a range of human phenotypes. We combined variants into a genetic score standardized to a 30% increase in relative activated partial thromboplastin time, equivalent to what can be achieved with pharmacological FXI reduction. Using data from 371 695 participants in the United Kingdom Biobank and 2 large-scale genome-wide association studies, we examined the effect of this FXI score on thrombotic and bleeding end points. Results- Genetic disposition to lower FXI levels was associated with reduced risks of venous thrombosis (odds ratio, 95% CI; P value; odds ratio=0.1, 0.07-0.14; P=3×10-43) and ischemic stroke (odds ratio=0.47, 0.36-0.61; P=2×10-8) but not with major bleeding (odds ratio=0.7, 0.45-1.04; P=0.0739). The observed relative risk reductions were consistent within a range of subgroups that were at high risk for thrombosis. Consistently, we observed higher absolute risk reductions conferred by genetically lower FXI levels in high-risk subgroups, such as patients with atrial fibrillation. Conclusions- Human genetic data suggest that pharmacological inhibition of FXI may achieve considerable reductions in ischemic stroke risk without clear evidence for an associated risk of major bleeding. The quantitative framework developed can be used to support the estimation of achievable risk reductions with pharmacological modulation of FXI.

The Assessment of Quality Attributes for Biosimilars: a Statistical Perspective on Current Practice and a Proposal.

Establishing comparability of the originator and its biosimilar at the structural and functional level, by analyzing so-called quality attributes, is an important step in biosimilar development. The statistical assessment of quality attributes is currently in the focus of attention because both the FDA and the EMA are working on regulatory documents for advising companies on the use of statistical approaches for strengthening their comparability claim. In this paper, we first discuss "comparable" and "not comparable" settings and propose a shift away from the usual comparison of the mean values: we argue that two products can be considered comparable if the range of the originator fully covers the range of the biosimilar. We then introduce a novel statistical testing procedure (the "tail-test") and compare the operating characteristics of the proposed approach with approaches currently used in practice. In contrast to the currently used approaches, we note that our proposed methodology is compatible with the proposed understanding of comparability and has, compared to other frequently applied range-based approaches, the advantage of being a formal statistical testing procedure which controls the patient's risk and has reasonable large-sample properties.

Incorporating historical information in biosimilar trials: Challenges and a hybrid Bayesian-frequentist approach.

For the approval of biosimilars, it is, in most cases, necessary to conduct large Phase III clinical trials in patients to convince the regulatory authorities that the product is comparable in terms of efficacy and safety to the originator product. As the originator product has already been studied in several trials beforehand, it seems natural to include this historical information into the showing of equivalent efficacy. Since all studies for the regulatory approval of biosimilars are confirmatory studies, it is required that the statistical approach has reasonable frequentist properties, most importantly, that the Type I error rate is controlled-at least in all scenarios that are realistic in practice. However, it is well known that the incorporation of historical information can lead to an inflation of the Type I error rate in the case of a conflict between the distribution of the historical data and the distribution of the trial data. We illustrate this issue and confirm, using the Bayesian robustified meta-analytic-predictive (MAP) approach as an example, that simultaneously controlling the Type I error rate over the complete parameter space and gaining power in comparison to a standard frequentist approach that only considers the data in the new study, is not possible. We propose a hybrid Bayesian-frequentist approach for binary endpoints that controls the Type I error rate in the neighborhood of the center of the prior distribution, while improving the power. We study the properties of this approach in an extensive simulation study and provide a real-world example.

An update on the clinical evidence that supports biosimilar approvals in Europe.

AIM: Sponsors and regulators have more than 10 years of experience with the development of biosimilars in Europe. However, the regulatory pathway is still evolving. The present article provides an update on biosimilar development in practice by reviewing the clinical development programmes of recently approved biosimilars in Europe. METHODS: We used the European public assessment reports (EPARs) which are published by the European Medicines Agency (EMA) for a comparison of the clinical development programmes of the 37 approved biosimilars in Europe. Here, we present novel strategies in the development of biosimilars by focusing specifically on the 17 biosimilars that have gained approval in the last year, but we also compare additional key characteristics for all approved biosimilars. RESULTS: The high variability of the clinical development strategies that we found previously was confirmed in the present analysis. Compared with earlier biosimilar applications, more nonstandard development strategies have been used recently. This includes, for example, applications without any studies in patients, and more complex study designs. During this study, we found that the EPARs for biosimilars seem to be improving; however, we identified important details which were still often missing. We provide a proposal for a checklist of the minimum information that should be included in biosimilar EPARs for giving the general public insights into the rationale for the approval of biosimilars. CONCLUSIONS: European regulators still seem to be open to consider approaches that differ from the guidelines or previous applications, as long as justification is provided.

Longitudinal assessment of the impact of multiple switches between a biosimilar and its reference product on efficacy parameters.

Patients, physicians, and health care providers in Europe have more than 10 years of experience with biosimilars. However, there are still debates if switching between a biosimilar and its reference product influences the efficacy of the treatment. In this paper, we address this uncertainty by developing a formal statistical test that can be used for showing that switching has no negative impact on the efficacy of biosimilars. For that, we first introduce a linear mixed-effects model that is used for defining the null hypothesis (switching influences the efficacy) and the alternative hypothesis (switching has no influence on the efficacy). Using this as the foundation of our work, we propose several approaches for testing for changes in the efficacy of the treatment due to switching and discuss the properties of these tests in an extensive simulation study. It is shown that all these methods have advantages and disadvantages and the decision regarding which method is preferred depends on the expectation of a switching assessment. To demonstrate the applicability of the methods in practice, the approaches were applied to the data of the EGALITY study, which compares the reference product Enbrel® (Amgen) with the approved biosimilar Erelzi® (Sandoz).

A Longitudinal Item Response Theory Model to Characterize Cognition Over Time in Elderly Subjects.

For drug development in neurodegenerative diseases such as Alzheimer's disease, it is important to understand which cognitive domains carry the most information on the earliest signs of cognitive decline, and which subject characteristics are associated with a faster decline. A longitudinal Item Response Theory (IRT) model was developed for the Basel Study on the Elderly, in which the Consortium to Establish a Registry for Alzheimer's Disease - Neuropsychological Assessment Battery (with additions) and the California Verbal Learning Test were measured on 1,750 elderly subjects for up to 13.9 years. The model jointly captured the multifaceted nature of cognition and its longitudinal trajectory. The word list learning and delayed recall tasks carried the most information. Greater age at baseline, fewer years of education, and positive APOEɛ4 carrier status were associated with a faster cognitive decline. Longitudinal IRT modeling is a powerful approach for progressive diseases with multifaceted endpoints.

Blood pressure control and cardiovascular risk in hypertensive patients with type 2 diabetes: The German T2Target registry.

Blood pressure control in patients with type 2 diabetes and hypertension is poor. It is unclear how effectively general practitioners assess and treat such patients. T2Target included hypertensive patients with type 2 diabetes mellitus who had undergone ambulatory blood pressure monitoring within the past 3 months. Recordings were analyzed by the general practitioner and an independent center and the conclusions were compared. Nighttime hypertension was reported less frequently by the general practitioner in comparison with central assessment (43.9% vs 77.9%, P<.001), as were masked (4.0% vs 13.1%, P<.001) and isolated office (4.4% vs 8.8%, P<.001) hypertension. A total of 13.9% of patients were deemed to have controlled blood pressure (ambulatory blood pressure monitoring). For the 784 patients with uncontrolled blood pressure, 40.7% underwent no change to their antihypertensive treatment. Cardiovascular risk was underestimated, with 11.1% deemed to be at very high risk, in contrast to the 97.0% of patients by central assessment. In conclusion, blood pressure control in hypertensive patients with type 2 diabetes mellitus is poor and not accurately assessed by office-based general practitioners, despite the use of ambulatory blood pressure monitoring.

Clinical trials for authorized biosimilars in the European Union: a systematic review.

AIM: In 2006, Omnitrope (by Sandoz) was the first approved biosimilar in Europe. To date, 21 biosimilars for seven different biologics are on the market. The present study compared the clinical trials undertaken to obtain market authorization. METHODS: We summarized the findings of a comprehensive review of all clinical trials up to market authorization of approved biosimilars, using the European public assessment reports (EPARs) published by the European Medicines Agency (EMA). The features compared were, among others, the number of patients enrolled, the number of trials, the types of trial design, choice of endpoints and equivalence margins for pharmacokinetic (PK)/pharmacodynamic (PD) and phase III trials. RESULTS: The variability between the clinical development strategies is high. Some differences are explainable by the characteristics of the product; if, for example, the PD marker can be assumed to predict the clinical outcome, no efficacy trials might be necessary. However, even for products with the same reference product, the sample size, endpoints and statistical models are not always the same. CONCLUSIONS: There seems to be flexibility for sponsors regarding the decision as to how best to prove biosimilarity.