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.

Sex differences in the generalizability of randomized clinical trials in heart failure with reduced ejection fraction.

AIMS: In order to understand how sex differences impact the generalizability of randomized clinical trials (RCTs) in patients with heart failure (HF) and reduced ejection fraction (HFrEF), we sought to compare clinical characteristics and clinical outcomes between RCTs and HF observational registries stratified by sex. METHODS AND RESULTS: Data from two HF registries and five HFrEF RCTs were used to create three subpopulations: one RCT population (n = 16 917; 21.7% females), registry patients eligible for RCT inclusion (n = 26 104; 31.8% females), and registry patients ineligible for RCT inclusion (n = 20 810; 30.2% females). Clinical endpoints included all-cause mortality, cardiovascular mortality, and first HF hospitalization at 1 year. Males and females were equally eligible for trial enrolment (56.9% of females and 55.1% of males in the registries). One-year mortality rates were 5.6%, 14.0%, and 28.6% for females and 6.9%, 10.7%, and 24.6% for males in the RCT, RCT-eligible, and RCT-ineligible groups, respectively. After adjusting for 11 HF prognostic variables, RCT females showed higher survival compared to RCT-eligible females (standardized mortality ratio [SMR] 0.72; 95% confidence interval [CI] 0.62-0.83), while RCT males showed higher adjusted mortality rates compared to RCT-eligible males (SMR 1.16; 95% CI 1.09-1.24). Similar results were also found for cardiovascular mortality (SMR 0.89; 95% CI 0.76-1.03 for females, SMR 1.43; 95% CI 1.33-1.53 for males). CONCLUSION: Generalizability of HFrEF RCTs differed substantially between the sexes, with females having lower trial participation and female trial participants having lower mortality rates compared to similar females in the registries, while males had higher than expected cardiovascular mortality rates in RCTs compared to similar males in registries.

Diagnostic and prognostic factors in patients with prostate cancer: a systematic review.

OBJECTIVES: As part of the PIONEER Consortium objectives, we have explored which diagnostic and prognostic factors (DPFs) are available in relation to our previously defined clinician and patient-reported outcomes for prostate cancer (PCa). DESIGN: We performed a systematic review to identify validated and non-validated studies. DATA SOURCES: MEDLINE, Embase and the Cochrane Library were searched on 21 January 2020. ELIGIBILITY CRITERIA: Only quantitative studies were included. Single studies with fewer than 50 participants, published before 2014 and looking at outcomes which are not prioritised in the PIONEER core outcome set were excluded. DATA EXTRACTION AND SYNTHESIS: After initial screening, we extracted data following the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of prognostic factor studies (CHARMS-PF) criteria and discussed the identified factors with a multidisciplinary expert group. The quality of the included papers was scored for applicability and risk of bias using validated tools such as PROBAST, Quality in Prognostic Studies and Quality Assessment of Diagnostic Accuracy Studies 2. RESULTS: The search identified 6604 studies, from which 489 DPFs were included. Sixty-four of those were internally or externally validated. However, only three studies on diagnostic and seven studies on prognostic factors had a low risk of bias and a low risk concerning applicability. CONCLUSION: Most of the DPFs identified require additional evaluation and validation in properly designed studies before they can be recommended for use in clinical practice. The PIONEER online search tool for DPFs for PCa will enable researchers to understand the quality of the current research and help them design future studies. ETHICS AND DISSEMINATION: There are no ethical implications.

Updating and Integrating Core Outcome Sets for Localised, Locally Advanced, Metastatic, and Nonmetastatic Castration-resistant Prostate Cancer: An Update from the PIONEER Consortium.

CONTEXT: Harmonisation of outcome reporting and definitions for clinical trials and routine patient records can enable health care systems to provide more efficient outcome-driven and patient-centred interventions. We report on the work of the PIONEER Consortium in this context for prostate cancer (PCa). OBJECTIVE: To update and integrate existing core outcome sets (COS) for PCa for the different stages of the disease, assess their applicability, and develop standardised definitions of prioritised outcomes. EVIDENCE ACQUISITION: We followed a four-stage process involving: (1) systematic reviews; (2) qualitative interviews; (3) expert group meetings to agree standardised terminologies; and (4) recommendations for the most appropriate definitions of clinician-reported outcomes. EVIDENCE SYNTHESIS: Following four systematic reviews, a multinational interview study, and expert group consensus meetings, we defined the most clinically suitable definitions for (1) COS for localised and locally advanced PCa and (2) COS for metastatic and nonmetastatic castration-resistant PCa. No new outcomes were identified in our COS for localised and locally advanced PCa. For our COS for metastatic and nonmetastatic castration-resistant PCa, nine new core outcomes were identified. CONCLUSIONS: These are the first COS for PCa for which the definitions of prioritised outcomes have been surveyed in a systematic, transparent, and replicable way. This is also the first time that outcome definitions across all prostate cancer COS have been agreed on by a multidisciplinary expert group and recommended for use in research and clinical practice. To limit heterogeneity across research, these COS should be recommended for future effectiveness trials, systematic reviews, guidelines and clinical practice of localised and metastatic PCa. PATIENT SUMMARY: Patient outcomes after treatment for prostate cancer (PCa) are difficult to compare because of variability. To allow better use of data from patients with PCa, the PIONEER Consortium has standardised and recommended outcomes (and their definitions) that should be collected as a minimum in all future studies.

Generalizability of randomized controlled trials in heart failure with reduced ejection fraction.

BACKGROUND: Heart failure (HF) trials have stringent inclusion and exclusion criteria, but limited data exist regarding generalizability of trials. We compared patient characteristics and outcomes between patients with HF and reduced ejection fraction (HFrEF) in trials and observational registries. METHODS AND RESULTS: Individual patient data for 16 922 patients from five randomized clinical trials and 46 914 patients from two HF registries were included. The registry patients were categorized into trial-eligible and non-eligible groups using the most commonly used inclusion and exclusion criteria. A total of 26 104 (56%) registry patients fulfilled the eligibility criteria. Unadjusted all-cause mortality rates at 1 year were lowest in the trial population (7%), followed by trial-eligible patients (12%) and trial-non-eligible registry patients (26%). After adjustment for age and sex, all-cause mortality rates were similar between trial participants and trial-eligible registry patients [standardized mortality ratio (SMR) 0.97; 95% confidence interval (CI) 0.92-1.03] but cardiovascular mortality was higher in trial participants (SMR 1.19; 1.12-1.27). After full case-mix adjustment, the SMR for cardiovascular mortality remained higher in the trials at 1.28 (1.20-1.37) compared to RCT-eligible registry patients. CONCLUSION: In contemporary HF registries, over half of HFrEF patients would have been eligible for trial enrolment. Crude clinical event rates were lower in the trials, but, after adjustment for case-mix, trial participants had similar rates of survival as registries. Despite this, they had about 30% higher cardiovascular mortality rates. Age and sex were the main drivers of differences in clinical outcomes between HF trials and observational HF registries.

Diagnostic and prognostic factors in patients with prostate cancer: a systematic review protocol.

INTRODUCTION: As part of the PIONEER (Prostate Cancer Diagnosis and Treatment Enhancement Through the Power of Big Data in Europe) Consortium, we will explore which diagnostic and prognostic factors (DPFs) are currently being researched to previously defined clinical and patient-reported outcomes for prostate cancer (PCa). METHODS AND ANALYSIS: This research project will follow the following four steps: (1) a broad systematic literature review of DPFs for all stages of PCa, covering evidence from 2014 onwards; (2) discussion of systematic review findings by a multidisciplinary expert panel; (3) risk of bias assessment and applicability with Prediction model Risk Of Bias Assessment Tool criteria, Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) and the Quality In Prognosis Studies tool (QUIPS) and (4) additional quantitative assessments if required. ETHICS AND DISSEMINATION: We aim to develop an online tool to present the DPFs identified in this research and make them available across all stakeholders. There are no ethical implications.

A Bayesian model to estimate the cutoff and the clinical utility of a biomarker assay.

To enable targeted therapies and enhance medical decision-making, biomarkers are increasingly used as screening and diagnostic tests. When using quantitative biomarkers for classification purposes, this often implies that an appropriate cutoff for the biomarker has to be determined and its clinical utility must be assessed. In the context of drug development, it is of interest how the probability of response changes with increasing values of the biomarker. Unlike sensitivity and specificity, predictive values are functions of the accuracy of the test, depend on the prevalence of the disease and therefore are a useful tool in this setting. In this paper, we propose a Bayesian method to not only estimate the cutoff value using the negative and positive predictive values, but also estimate the uncertainty around this estimate. Using Bayesian inference allows us to incorporate prior information, and obtain posterior estimates and credible intervals for the cut-off and associated predictive values. The performance of the Bayesian approach is compared with alternative methods via simulation studies of bias, interval coverage and width and illustrations on real data with binary and time-to-event outcomes are provided.

Model selection based on combined penalties for biomarker identification.

The growing role of targeted medicine has led to an increased focus on the development of actionable biomarkers. Current penalized selection methods that are used to identify biomarker panels for classification in high-dimensional data, however, often result in highly complex panels that need careful pruning for practical use. In the framework of regularization methods, a penalty that is a weighted sum of the L1 and L0 norm has been proposed to account for the complexity of the resulting model. In practice, the limitation of this penalty is that the objective function is non-convex, non-smooth, the optimization is computationally intensive and the application to high-dimensional settings is challenging. In this paper, we propose a stepwise forward variable selection method which combines the L0 with L1 or L2 norms. The penalized likelihood criterion that is used in the stepwise selection procedure results in more parsimonious models, keeping only the most relevant features. Simulation results and a real application show that our approach exhibits a comparable performance with common selection methods with respect to the prediction performance while minimizing the number of variables in the selected model resulting in a more parsimonious model as desired.

A Bayesian Approach for Analysis of Whole-Genome Bisulfite Sequencing Data Identifies Disease-Associated Changes in DNA Methylation.

DNA methylation is a key epigenetic modification involved in gene regulation whose contribution to disease susceptibility remains to be fully understood. Here, we present a novel Bayesian smoothing approach (called ABBA) to detect differentially methylated regions (DMRs) from whole-genome bisulfite sequencing (WGBS). We also show how this approach can be leveraged to identify disease-associated changes in DNA methylation, suggesting mechanisms through which these alterations might affect disease. From a data modeling perspective, ABBA has the distinctive feature of automatically adapting to different correlation structures in CpG methylation levels across the genome while taking into account the distance between CpG sites as a covariate. Our simulation study shows that ABBA has greater power to detect DMRs than existing methods, providing an accurate identification of DMRs in the large majority of simulated cases. To empirically demonstrate the method's efficacy in generating biological hypotheses, we performed WGBS of primary macrophages derived from an experimental rat system of glomerulonephritis and used ABBA to identify >1000 disease-associated DMRs. Investigation of these DMRs revealed differential DNA methylation localized to a 600 bp region in the promoter of the Ifitm3 gene. This was confirmed by ChIP-seq and RNA-seq analyses, showing differential transcription factor binding at the Ifitm3 promoter by JunD (an established determinant of glomerulonephritis), and a consistent change in Ifitm3 expression. Our ABBA analysis allowed us to propose a new role for Ifitm3 in the pathogenesis of glomerulonephritis via a mechanism involving promoter hypermethylation that is associated with Ifitm3 repression in the rat strain susceptible to glomerulonephritis.

Nationwide and population-based prescription patterns in bipolar disorder.

OBJECTIVES: The aim of the present study was to describe prescription patterns and changes in these patterns over the last decade for patients diagnosed with bipolar disorder in mental healthcare, using population-based and nationwide data, and to relate the findings to recommendations from international guidelines. METHODS: A population-based, nationwide study was carried out. It included register-based longitudinal data on all patients with a first-ever contact with mental healthcare with a diagnosis of mania/bipolar disorder from the entire Danish population, and all prescription data for this population during the decade from 2000 to 2011, inclusive. RESULTS: A total of 3,205 patients were included in the study. Lithium was prescribed less, and antiepileptic and atypical antipsychotic agents were prescribed substantially more during the study period. Lithium went from being the first drug prescribed to being the last, and was replaced by atypical antipsychotic agents. Antiepileptic agents went from being the fourth to the second drug class prescribed, and the prescription of antidepressants was virtually unchanged, at a high level, during the decade (one-year value 40-60%). The prescription of lamotrigine and quetiapine increased substantially. Combination therapy increased for all drug combinations, except for lithium combined with antidepressants. CONCLUSIONS: Major changes took place in drug prescriptions during the study period. The decrease in the use of lithium and the constant high use of antidepressants do not align with recommendations from international guidelines.

The Association of Reproductive Hormone Levels and All-Cause, Cancer, and Cardiovascular Disease Mortality in Men.

CONTEXT: Testosterone (T) levels have been associated with mortality, but controversy exists. OBJECTIVE: Our objective was to investigate associations between serum levels of total T, SHBG, free T, estradiol, LH and FSH, and subsequent mortality with up to 30 years of follow-up. DESIGN: This was a prospective cohort study consisting of men participating in four independent population-based surveys (MONICA I-III and Inter99) from 1982 to 2001 and followed until December 2012 with complete registry follow-up. SETTING AND PARTICIPANTS: A total of 5350 randomly selected men from the general population aged 30, 40, 50, 60, or 70 years at baseline participated. MAIN OUTCOMES AND MEASURES: All-cause mortality, cardiovascular disease (CVD) mortality, and cancer mortality were the main outcomes. RESULTS: A total of 1533 men died during the follow-up period; 428 from CVD and 480 from cancer. Cox proportional hazard models revealed that men in highest LH quartile had an increased all-cause mortality compared to lowest quartile (hazard ratio [HR], 1.32; 95% confidence interval [CI], 1.14-1.53). Likewise, increased quartiles of LH/T and estradiol increased the risk of all-cause mortality (HR, 1.23; 95% CI, 1.06-1.43; HR, 1.23; 95% CI 1.06-1.43). No association to T levels was found. Higher LH levels were associated with increased cancer mortality (HR, 1.42; 95% CI, 1.10-1.84) independently of smoking status. Lower CVD mortality was seen for men with T in the highest quartile compared to lowest (HR, 0.72; 95% CI, 0.53-0.98). Furthermore, negative trends were seen for SHBG and free T in relation to CVD mortality, however insignificant. CONCLUSION: The observed positive association of LH and LH/T, but not T, with all-cause mortality suggests that a compensated impaired Leydig cell function may be a risk factor for death by all causes in men. Our findings underpin the clinical importance of including LH measurement in the diagnostic work-up of male patients seeking help for possible androgen insufficiency.

Causes of decreased life expectancy over the life span in bipolar disorder.

BACKGROUND: Accelerated aging has been proposed as a mechanism explaining the increased prevalence of comorbid general medical illnesses in bipolar disorder. AIMS: To test the hypothesis that lost life years due to natural causes starts in early and mid-adulthood, supporting the hypothesis of accelerated aging. METHODS: Using individual data from nationwide registers of patient with a diagnosis of bipolar disorder we calculated remaining life expectancies before age 90 years for values of age 15, 25, 35…75 years among all individuals alive in year 2000. Further, we estimated the reduction in life expectancy due to natural causes (physical illnesses) and unnatural causes (suicide and accidents) in relation to age. RESULTS: A total of 22,635 patients with bipolar disorder were included in the study in addition to data from the entire Danish general population of 5.4 million people. At age 15 years, remaining life expectancy before age 90 years was decreased 12.7 and 8.9 life years, respectively, for men and women with bipolar disorder. For 15-year old boys with bipolar disorder, natural causes accounted for 58% of all lost life years and for 15-year old girls, natural causes accounted for 67% increasing to 74% and 80% for 45-year old men and women, respectively. LIMITATIONS: Data concern patients who get contact to hospital psychiatry only. CONCLUSIONS: Natural causes of death is the most prevalent reason for lost life years already from adolescence and increases substantially during early and mid-adulthood, in this way supporting the hypothesis of accelerated aging. Early intervention in bipolar disorder should not only focus on improving outcome of the bipolar disorder but also on decreasing the risk of comorbid general medical illnesses.

Life expectancy in bipolar disorder.

OBJECTIVE: Life expectancy in patients with bipolar disorder has been reported to be decreased by 11 to 20 years. These calculations are based on data for individuals at the age of 15 years. However, this may be misleading for patients with bipolar disorder in general as most patients have a later onset of illness. The aim of the present study was to calculate the remaining life expectancy for patients of different ages with a diagnosis of bipolar disorder. METHODS: Using nationwide registers of all inpatient and outpatient contacts to all psychiatric hospitals in Denmark from 1970 to 2012 we calculated remaining life expectancies for values of age 15, 25, 35 ⃛ 75 years among all individuals alive in year 2000. RESULTS: For the typical male or female patient aged 25 to 45 years, the remaining life expectancy was decreased by 12.0-8.7 years and 10.6-8.3 years, respectively. The ratio between remaining life expectancy in bipolar disorder and that of the general population decreased with age, indicating that patients with bipolar disorder start losing life-years during early and mid-adulthood. CONCLUSIONS: Life expectancy in bipolar disorder is decreased substantially, but less so than previously reported. Patients start losing life-years during early and mid-adulthood.

Diagnostic stability in pediatric bipolar disorder.

BACKGROUND: The diagnostic stability of pediatric bipolar disorder has not been investigated previously. The aim was to investigate the diagnostic stability of the ICD-10 diagnosis of pediatric mania/bipolar disorder. METHODS: All patients below 19 years of age who got a diagnosis of mania/bipolar disorder at least once in a period from 1994 to 2012 at psychiatric inpatient or outpatient contact in Denmark were identified in a nationwide register. RESULTS: Totally, 354 children and adolescents got a diagnosis of mania/bipolar disorder at least once; a minority, 144 patients (40.7%) got the diagnosis at the first contact whereas the remaining patients (210; 59.3%) got the diagnosis at later contacts before age 19. For the latter patients, the median time elapsed from first treatment contact with the psychiatric service system to the first diagnosis with a manic episode/bipolar disorder was nearly 1 year and for 25% of those patients it took more than 2½ years before the diagnosis was made. The most prevalent other diagnoses than bipolar disorder at first contact were depressive disorder (21.4%), acute and transient psychotic disorders or other non-organic psychosis (19.2%), reaction to stress or adjustment disorder (14.8%) and behavioral and emotional disorders with onset during childhood or adolescents (10.9%). Prevalence rates of schizophrenia, personality disorders, anxiety disorder or hyperkinetic disorders (ADHD) were low. LIMITATIONS: Data concern patients who get contact to hospital psychiatry only. CONCLUSIONS: Clinicians should be more observant on manic symptoms in children and adolescents who at first glance present with transient psychosis, reaction to stress/adjustment disorder or with behavioral and emotional disorders with onset during childhood or adolescents (F90-98) and follow these patients more closely over time identifying putable hypomanic and manic symptoms as early as possible.

Are rates of pediatric bipolar disorder increasing? Results from a nationwide register study.

Studies from the USA suggest that rates of pediatric bipolar disorder have increased since the mid-90s, but no study outside the USA has been published on the rates of pediatric bipolar disorder. Further, it is unclear whether an increase in rates reflects a true increase in the illness or more diagnostic attention. Using nationwide registers of all inpatients and outpatients contacts to all psychiatric hospitals in Denmark, we investigated (1) gender-specific rates of incident pediatric mania/bipolar disorder during a period from 1995 to 2012, (2) whether age and other characteristics for pediatric mania/bipolar disorder changed during the calendar period (1995 to 2003 versus 2004 to 2012), and (3) whether the diagnosis is more often made at first psychiatric contact in recent time compared to earlier according to gender. Totally, 346 patients got a main diagnosis of a manic episode (F30) or bipolar affective disorder (F31) at least once during the study period from 1995 to 2012. For both sexes, annual rates of mania/bipolar disorder two to four doubled during the study period (0.001% before year 2004 to 0.002%-0.004% in 2010). Median age at the index diagnosis was very similar during the two calendar periods (17.2, quartiles, 16.2-18.3 versus 17.4, quartiles, 16.1-18.2) indicating that the diagnosis of mania/bipolar disorder was not made earlier in the recent calendar period. Similarly, there were no differences between early versus late in the study period in the fractions of first contact diagnosis of mania/bipolar disorder diagnoses, the contact number at which patients got the diagnosis or the duration from first psychiatric contact to the diagnosis of mania/bipolar disorder. The rate of diagnosis of mania/bipolar disorder increased from 1995 to 2014, which did not seem to be explained by more diagnostic attention.

Starting lithium prophylaxis early v. late in bipolar disorder.

BACKGROUND: No study has investigated when preventive treatment with lithium should be initiated in bipolar disorder. AIMS: To compare response rates among patients with bipolar disorder starting treatment with lithium early v. late. METHOD: Nationwide registers were used to identify all patients with a diagnosis of bipolar disorder in psychiatric hospital settings who were prescribed lithium during the period 1995-2012 in Denmark (n = 4714). Lithium responders were defined as patients who, following a stabilisation lithium start-up period of 6 months, continued lithium monotherapy without being admitted to hospital. Early v. late intervention was defined in two ways: (a) start of lithium following first contact; and (b) start of lithium following a diagnosis of a single manic/mixed episode. RESULTS: Regardless of the definition used, patients who started lithium early had significantly decreased rates of non-response to lithium compared with the rate for patients starting lithium later (adjusted analyses: first v. later contact: P<0.0001; hazard ratio (HR) = 0.87, 95% CI 0.76-0.91; single manic/mixed episode v. bipolar disorder: P<0.0001; HR = 0.75, 95% CI 0.67-0.84). CONCLUSIONS: Starting lithium treatment early following first psychiatric contact or a single manic/mixed episode is associated with increased probability of lithium response.