Erzsébet Horváth-Puhó

Real‐world data are increasingly important for regulatory and health technology assessment decision‐making, yet their effective use across Europe remains limited by technical, methodological, regulatory, and societal challenges. Drawing on the experiences of six Horizon Europe projects in the MetReal cluster, this review summarizes common barriers and emerging solutions for the use of real‐world data in regulatory decision‐making. Key challenges include complex and lengthy data access procedures, fragmented governance frameworks, variable data quality and completeness, limited interoperability, methodological difficulties in analyzing heterogeneous datasets, and infrastructure constraints within secure data environments. The article highlights opportunities to address these barriers through improved metadata, harmonization, federated analytics, synthetic data sandboxes, sustainable funding, and stronger collaboration between researchers, regulators, health technology assessment bodies, data holders, and citizens. These recommendations aim to strengthen Europe's health data ecosystem and support more robust, efficient, and trusted evidence generation for public health and patient benefit.

Background: There is increasing recognition that trauma exposure and related psychiatric consequences predict cardiovascular disease risk. However, most research has specifically examined post‐traumatic stress disorder—despite evidence that a range of psychiatric disorders may follow trauma. We applied machine learning to identify key post‐traumatic psychiatric predictors of incident major adverse cardiac and cerebrovascular events (MACCE) in a population‐based cohort. Methods: We conducted a case–control study of adults aged ≥18 years, nested within a trauma‐exposed cohort of all Denmark residents from 1995 to 2018. Cases ( n =43 994) experienced MACCE (non‐fatal myocardial infarction, non‐fatal stroke, coronary revascularization, cardiovascular death) between 1995 and 2019. Controls ( n =175 872) were matched 4:1 on MACCE date, sex, year of trauma, and age at trauma. Machine learning model predictors included a range of psychiatric disorders diagnosed ≤3 years following trauma. Sex‐stratified machine learning models were trained using extreme gradient boosting (men: n =91 111; women: n =84 781) and tested in hold‐out datasets (men: n =22 777; women: n =21 197). We identified the 15 most important psychiatric disorder predictors of MACCE using Shapley Additive Explanations values. Results: For men and women, the most important psychiatric predictor of MACCE was alcohol use disorders. Other important psychiatric predictors included depressive disorders, post‐traumatic stress disorder/stress‐related disorders, brain‐related/physiological conditions, and non‐substance‐induced delirium. Some sex‐specific findings also emerged (eg, depressive episodes more important for women than men). Overall, odds ratios in training and testing sets indicated that odds of MACCE were greater among those with (versus without) psychiatric disorders. Conclusions: Results emphasize the importance of considering a range of psychiatric disorders when assessing cardiovascular risk following trauma.

Objective: To assess whether early-life antibiotics are associated with spondyloarthritis (SpA) diagnosed by age 21. Methods: In this population-based, matched case-control study, we used Danish live births (1997-2023) restricted to those reaching the following minimum relevant ages by December 31, 2024: 1-21 years for psoriatic arthritis (PsA), and 6-21 years for peripheral/axial SpA and inflammatory bowel disease (IBD)-associated arthritis. Cases (n = 560) had physician-recorded SpA (peripheral/axial SpA, PsA, or IBD-associated arthritis). Controls were matched 1:50 by birth year, sex, and calendar year of diagnosis. Primary exposure was systemic antibiotic use the first year of life; secondary exposures included antibiotic class, number of courses, delivery mode, early upper respiratory tract infection (URTI), and systemic nonbacterial antimicrobials. Conditional logistic regression estimated adjusted odds ratios (aORs); dose response was tested with the Wald test. Results: Among cases (median age 16.8 years; 59% peripheral/axial SpA, 28% PsA, 13% IBD-associated arthritis), 50% received antibiotics, vs 42% of controls. First-year antibiotic exposure was associated with higher odds of SpA (aOR 1.35, 95% CI 1.14-1.59). Broad-spectrum penicillins had the strongest class-specific association (aOR 1.42, 95% CI 1.19-1.70). Early URTI was independently associated with SpA (aOR 1.88, 95% CI 1.32-2.67); delivery mode and systemic nonbacterial antimicrobials were not. In SpA subgroup analysis, the association with first-year antibiotics was significant only for PsA (aOR 1.75, 95% CI 1.27-2.41). Conclusion: Antibiotic exposure in the first year of life, particularly broad-spectrum penicillins, was associated with increased odds of SpA by age 21, supporting the hypothesis that early-life microbiome disruption may increase later SpA risk.

The growth in evidence-based medicine clearly benefits patient care but it is important that evidence is accessible and usable. Translating research results into usable, meaningful information can be challenging and impede the implementation of robust evidence. There are many aspects to making the statistical parts of a research study actionable for practicing clinicians and policy makers. These include ensuring that the right questions are asked, the study is designed appropriately, and results are transparent and applicable to the clinical setting. In this perspectives-style article we offer guidance on these considerations by highlighting several approaches that we have found effective for improving the interpretability and practical application of statistical findings. Our over-arching aim is to stimulate interdisciplinary dialogue throughout the research process. Specifically, we discuss the interpretation of p-values, effect estimates, differences between means, scaling regression coefficients, unadjusted/adjusted estimates, Minimal Clinically Important Difference, absolute and relative risk, and suggest how clinical meaning can be enhanced by presenting the same information in different but complementary ways. We conclude with a recommendation that study teams prioritize interdisciplinary discussions around clinical meaningfulness throughout our research studies to maximize their clinical impact.

Purpose: Cancer outcome studies frequently utilize registry data, which provide large-scale population-level information. However, these registries often lack detailed clinical information regarding comorbidities, lifestyle factors, and in-hospital treatments. The Central Denmark Cancer Cohort (CDCC) was established to address these limitations by linking medical and administrative registry data with clinical electronic medical records (EMR), to facilitate research to better understand the clinical course of cancer and its complications. Patients and Methods: The CDCC includes all patients with incident cancer, except non-melanoma skin cancer, diagnosed in the Central Denmark Region between 2012 and 2021, with complete follow-up through December 31, 2021. The CDCC was identified from the Danish Cancer Registry and linked via the Civil Personal Registration number to the Central Denmark Region Clinical Information System (CDRCIS) and to national registries, including the Danish National Patient Registry, National Prescription Registry, and Register of Laboratory Results for Research. We extracted data on demographics, lifestyle factors, comorbidities, treatments, and survival outcomes and assessed the availability of these data. Results: The CDCC included 68,028 patients with a median age of 68 years (interquartile range: 59– 76 years); 47.2% were female. The most common cancers were prostate cancer (10,024 patients), breast cancer (9,100 patients), and non-small cell lung carcinoma (7,264 patients). At diagnosis, 98.0% of patients had laboratory test results available, 87.5% received at least one in-hospital medication, and 86.9% had at least one characteristic documented in CDRCIS. During median follow-up of 2.6 years, 37.8% of patients died and five-year survival was 59.0% (95% confidence interval: 58.6– 59.4%). Data completeness varied by cancer type. Conclusion: The CDCC integrates clinical, lifestyle, and laboratory data with cancer registry information. By capturing factors typically unavailable in registry-based research, this platform offers a unique foundation for longitudinal studies on the clinical course of cancer.

Group B Streptococcus (GBS) is a common gut coloniser, yet it can cause serious disease in pregnant women, newborns, and in adults with predisposing conditions. Estimates suggest that approximately 20 million pregnant women are colonised with GBS, and 394,000 cases of invasive disease occur in infants every year, but there are many data gaps. Several countries lack robust data, leading to likely underestimation of the true disease burden, notably for GBS-associated stillbirths and preterm births. Recent progress in the development of maternal GBS vaccines adds urgency to the need for comprehensive surveillance data. The aim of this study is to describe advances in routine health information systems and record linkage offer new opportunities for robust monitoring and research on GBS outcomes, both before and after vaccine roll-out. This review article identifies crucial research gaps, proposes priority outcomes that should be monitored, and highlights opportunities to use routine data from three European countries (Denmark, the Netherlands, the United Kingdom) to harmonise data and strengthen the evidence base to inform national GBS vaccine policy decisions. Strengthening and harmonising routine data collection across Europe and globally will be essential to maximize and measure the benefits of a maternal GBS vaccination.

Response to Letter in regards to An Introduction to Competing Risks in Epidemiology [Letter]

Purpose: Mothers of children with epilepsy (MCE) face elevated risks of psychological stress and psychiatric disorders, yet their overall health care utilization post-diagnosis remains underexplored. Understanding these patterns may inform more integrated, evidence-based care for affected families. Patients and methods: A population-based cohort study using national health registry data in Denmark. From 1990 to 2017, 18,775 MCE were matched on maternal age, gestational age, child’s sex and age at epilepsy diagnosis with up to 10 mothers without a child diagnosed with epilepsy (n = 188,630) and followed until 2018. Period prevalence was calculated as the proportion of MCE health care utilization (HCU). We compared period prevalence of health care utilization (HCU)—including primary care, inpatient, outpatient, psychiatric, and surgical contacts—between MCE and matched comparators using modified Poisson regression to estimate risk ratios (RRs) and 95% confidence intervals (CIs). Mean annual HCU was visualized, and rate ratios were derived via negative binomial regression. Analyses were stratified by follow-up periods: 0– 5, 6– 10, and 11– 20 years post-index. Results: Over 20-years’ follow-up, MCE showed higher HCU than comparators across all services except primary care, which remained comparable. The RR of outpatient visits and surgical care contacts were slightly increased over the follow-up period. The risk of using psychiatric services was higher for MCE compared to the comparison cohort and increased over time, being 35% higher during the period of 11– 20 years of follow-up (RR_{11-20 years} = 1.35 [95% CI 1.18– 1.39]). The rate of HCU was increased for MCE compared to the comparison cohort across services, except for psychiatric services, which was comparable for MCE and the reference cohort. Conclusion: Mothers of children with epilepsy had more health care use across hospital services compared with matched mothers of children without epilepsy, regardless of the child’s age at diagnosis. These findings underscore the need for proactive mental health support and integrated care planning for mothers to mitigate long-term psychological burden.