European Journal Of Epidemiology[JOURNAL]
Sun
200 papers
Statistical methods for retrospective harmonization of longitudinal epidemiological data: a scoping review.
Zhang J, Behrendt J, Schultz T … +4 more , Aleksandrova K, Iqbal K, Pigeot I, Börnhorst C
Eur J Epidemiol · 2026 May · PMID 42207414 · Publisher ↗
Data harmonization is a prerequisite for joint cohort analyses. In this review, we aim to identify and contrast statistical methods for retrospective harmonization of longitudinal data. We performed a scoping review foll...
Data harmonization is a prerequisite for joint cohort analyses. In this review, we aim to identify and contrast statistical methods for retrospective harmonization of longitudinal data. We performed a scoping review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews guidelines. Studies were included if they described statistical methods for retrospectively harmonizing longitudinal data at the participant level. From 35 included papers out of 1,234 hits, we identified three types of statistical methods applicable to tabular data commonly collected in longitudinal epidemiological studies (e.g., questionnaires): (1) distribution-based methods, (2) the proportion score model, and (3) latent variable models. Our results suggest that the suitability of a statistical harmonization method mainly depends on the measurement scales of the original variables as well as on the type of target variable (directly measurable vs. latent). The chosen harmonization method influences how missing subsets of variables are addressed. None of the included studies applied more automated approaches such as machine learning-based procedures for deriving a harmonized dataset. Based on our findings, we present a roadmap that can guide researchers in selecting the most appropriate statistical method for a specific harmonization task and in handling variables collected only in a subset of studies. Data harmonization is still a demanding task that requires the development and application of novel tools for automating the procedures.
The Danish Infection Cohort: a resource for population based infectious disease epidemiology.
Cocoros NM, Oren E, Darvalics B … +5 more , Mortensen K, Jensen HAR, Eliasen MH, Pedersen L, Sørensen HT
Eur J Epidemiol · 2026 May · PMID 42201459 · Publisher ↗
Infections are a major cause of morbidity and mortality, but many aspects of the epidemiology of infections are unknown or unclear because key data are lacking. We created The Danish Infection Cohort by linking Danish Na...
Infections are a major cause of morbidity and mortality, but many aspects of the epidemiology of infections are unknown or unclear because key data are lacking. We created The Danish Infection Cohort by linking Danish National Health Survey data (including self-reported behavioral, health, and other information) to data on diagnoses and medications in Danish national health registries. We included individuals ≥ 18 years of age in surveys from 2010, 2013, 2017, and 2021. We identified evidence of infection in the 365 days post-survey by using two definitions: hospital-diagnosed infections, via primary or secondary discharge diagnoses in the Danish National Patient Registry (containing all Danish hospital-based outpatient, emergency, and inpatient encounters); and community-treated infections, via systemic anti-infective prescription redemptions at community pharmacies in the Danish National Prescription Registry. Overall, among 609,224 individuals, 196,980 (32.3%) had one or more infections within 365 days post-survey: 11,850 hospital-diagnosed and 185,130 community-treated. Because more than one infection occurred in some individuals, the total number of infections was 315,689: 25,385 hospital-diagnosed and 290,304 community-treated. The most common diagnoses were respiratory infections (34.9%), and the most common anti-infectives were antibiotics (80.6%). This cohort, with nearly 200,000 individuals with infections, and more than 315,000 infections overall, includes data on smoking, weight, physical activity, diet, alcohol use, and other factors, enabling a wide range of investigations including infectious disease patterns and etiologies over time.
The analysis of cancer screening trials in which the outcome is the incidence of late-stage disease.
Weiss NS, Kessler L, Etzioni R
Eur J Epidemiol · 2026 Apr · PMID 42201458 · Publisher ↗
The National Health Service-Galleri randomized trial seeks to determine the extent to which a blood-based multi-cancer screening test can lead to a reduced incidence of a variety of cancers that are at a late stage as of...
The National Health Service-Galleri randomized trial seeks to determine the extent to which a blood-based multi-cancer screening test can lead to a reduced incidence of a variety of cancers that are at a late stage as of the time of initial diagnosis. In February 2026, the investigators of this trial issued a news release indicating that after several years of follow-up, for a group of 12 pre-specified forms of cancer there was a lower incidence of stage IV cancer among patients in the intervention arm of the trial than in the control arm. The release stated that a group of participants whose stage IV cancer could not have been influenced by the screening - those with stage IV disease already present as of the initial screening exam - had been excluded from this analysis. We argue that while it is appropriate to restrict such an analysis to persons without prevalent stage IV cancer at the start of the trial, a valid result will be obtained only after accounting for the counterparts of these cases that are present in the trial's control arm. We describe how this can be accomplished, and the assumptions needed for such a strategy to succeed.
The decade of double-trouble: live birth and stillbirth sex ratio variation in Finland during the COVID-19 pandemic and Russia's attack on Ukraine.
Helle S, Tanskanen AO, Danielsbacka M
Eur J Epidemiol · 2026 May · PMID 42201457 · Publisher ↗
Exogenous shocks experienced by the population like armed conflicts, natural disasters and economic downturns have all been suggested to correlate with variation in human sex ratio at birth, either through sex-specific e...
Exogenous shocks experienced by the population like armed conflicts, natural disasters and economic downturns have all been suggested to correlate with variation in human sex ratio at birth, either through sex-specific embryonic mortality and/or primary sex adjustment. Recently experienced COVID-19 pandemic has also been related to changes in birth sex ratio in some countries but the results have been mixed. In the aftermath of the pandemic, Russia's attack on Ukraine caused further emotional anxiety and economic hardship in Europe. This was also true in Finland, which has had a decade-long close trading history with its neighboring state, Russia. Here, we examined the influence of this "double trouble" on variation in proportion of males born alive as well as male proportion of stillbirths in Finland during 2000-2024, using high-quality monthly census panel data covering the entire Finnish population (a total of 1,355,037 live births and 4,096 stillbirths). The results from Bayesian aggregated binomial time-series regressions showed no evidence for COVID-19-related associations on either the proportion of male live births or stillbirths. We did find an increased proportion of males born alive after 4 months of Russia's offensive but this finding should be considered as exploratory rather than confirmatory. If proportion of male live births or stillbirths can be regarded as a health indicator of the population, our results suggest that Finland as a society showed marked resilience in terms of non-responsive sex ratios when facing these two adverse exogenous shocks.
Risk prediction for individual patients and the pitfalls of selecting an optimal prediction model: do not judge a model by its c-statistic.
van Dalen JW
Eur J Epidemiol · 2026 Apr · PMID 42201456 · Full text
Researchers often report the area under the curve (AUC/c-statistic), Brier-index (BI), and explained variation (R) to assess risk prediction model performance. This may be inappropriate for scores to estimate individual...
Researchers often report the area under the curve (AUC/c-statistic), Brier-index (BI), and explained variation (R) to assess risk prediction model performance. This may be inappropriate for scores to estimate individual patient risks because population characteristics like risk score distributions may influence these performance metrics. This study assessed performance values for a fully accurate risk score under different distributions. A risk score ranging 0-100% was simulated for 100,000 individuals with 1000 bootstraps under six distributions: normal, uniform, bimodal, extreme bimodal, uniform ascending and uniform descending. Outcomes were simulated with full accuracy: 1% of individuals with a score of 0.01 had an outcome, 2% with 0.02 etc. Values for the AUC, BI, BI scaled (BS), and R were calculated, and calibration plots made. Sensitivity analyses included narrowing risk score range and reducing score granularity. For a normally distributed score, performance indices did not exceed values generally considered poor (AUC = 0.67, BI = 0.23, BS = 0.08, R = 8%). Values were best for an extreme bimodal distribution (AUC = 0.93, BI = 0.10, BS = 0.61, R = 61%). Performance indices for the other distributions were moderate (AUC = 0.80-0.83, BI = 0.17-0.18, BS = 0.25-0.33, R = 25-33%). Calibration plots were perfect for all distributions. Values worsened with narrowing score ranges. Reducing granularity had marginal effects. Risk prediction model performance values depend strongly on risk distributions. They may not exceed values generally considered poor/moderate in many common risk distributions. When assessing patient risk prediction models, investigators and clinicians interested in the accuracy of risk estimations may prioritize calibration and net benefits over commonly used metrics like the AUC and BI.
Adiposity measures, body composition, and all-cause and cause-specific mortality in the Korean national health screening cohort: beyond body mass index.
Kim JE, Lee DH, Giovannucci E … +1 more , Oh H
Eur J Epidemiol · 2026 May · PMID 42154432 · Publisher ↗
While body mass index (BMI) is the most widely-used measure of adiposity, alternative measures, including waist circumference (WC), waist-to-height ratio (WHtR), weight-adjusted waist index (WWI), and a body shape index...
While body mass index (BMI) is the most widely-used measure of adiposity, alternative measures, including waist circumference (WC), waist-to-height ratio (WHtR), weight-adjusted waist index (WWI), and a body shape index (ABSI), have been proposed to reflect body composition. However, the dose-response relationships between these alternative measures and mortality have not been fully established. We conducted a prospective cohort analysis using data from 328,084 Korean adults (aged ≥ 40 years) who participated in the national health screening program. Using Cox proportional hazards models, we examined the associations between multiple adiposity measures (BMI, WC, WHtR, WWI, ABSI) and all-cause and cause-specific mortality, with extensive control for methodological biases. During follow-up (mean: 10.2 years), 11,580 deaths were identified. BMI, WC, and WHtR showed J-shaped associations with all-cause, cardiovascular, and cancer mortality. The J-shaped associations became monotonically positive after adjustment for fat-free mass index (FFMI). WWI and ABSI showed monotonically positive associations with all-cause and cardiovascular mortality regardless of adjustment for FFMI, which remained unchanged after further control for methodological issues. All adiposity measures under study were positively associated with the proportion of "high fat-free/high fat mass" body type. However, with the proportion of "high fat-free/low fat mass" body type, WWI and ABSI were inversely associated, whereas BMI, WC, and WHtR presented inverse U-shaped associations. After accounting for fat-free mass, adiposity measures were associated with increased mortality among Korean adults in a monotonic dose-response manner. Our findings suggest that WWI and ABSI may reflect unhealthy body composition and could serve as alternative adiposity measures for clinical risk assessment.
Combining external placebo-arm data in a prevention trial: effects of cocoa extract supplementation on cardiovascular disease.
Hamaya R, Li S, Rist PM … +3 more , Shadyab AH, Manson JE, Sesso HD
Eur J Epidemiol · 2026 May · PMID 42149390 · Publisher ↗
Rigorous integration of external controls in large-scale cardiovascular prevention randomized controlled trials (RCTs) is under-explored. The Cocoa Supplement and Multivitamin Outcomes Study (COSMOS), a primary preventio...
Rigorous integration of external controls in large-scale cardiovascular prevention randomized controlled trials (RCTs) is under-explored. The Cocoa Supplement and Multivitamin Outcomes Study (COSMOS), a primary prevention RCT of cocoa extract supplementation, initially showed a statistically non-significant effect on total cardiovascular disease (CVD). We aimed to obtain more robust estimates of cocoa extract effects by integrating external control data from The VITamin D and OmegA-3 TriaL (VITAL), a similarly designed large-scale RCT. We analyzed 21,442 COSMOS participants (median 3.6-year follow-up) and used the VITAL omega-3 placebo arm as an external control. Outcomes included the original COSMOS primary CVD endpoint (a composite of myocardial infarction [MI], stroke, cardiovascular death, and coronary revascularization) and major adverse cardiovascular events (MACE; MI, stroke, and cardiovascular death). Two statistical approaches were used: a propensity score-based inverse probability weighting (IPW)-weighted Cox proportional hazard model and a doubly robust method to estimate 4-year risk ratios (RRs) and 95% confidence intervals (CIs) via bootstrapping. IPW-weighted VITAL-placebo data showed similar covariate distributions and 4-year event rates to COSMOS-placebo. While COSMOS data alone showed non-significant RRs (CVD: 0.91 [95% CI 0.79, 1.04]; MACE: 0.84 [0.70, 1.00]), incorporating VITAL-placebo showed protective effects of cocoa extract for cardiovascular endpoints ; RRs were 0.87 [0.76, 0.98] for CVD and 0.80 [0.68, 0.93] for MACE in Cox models; and those were 0.84 [0.74, 0.95] for CVD and 0.77 [0.67, 0.89] for MACE in the doubly robust method. E-values and sensitivity analyses indicated robustness. The present results suggest that original non-significant results of COSMOS could be due to type 2 error. While additional assumptions are needed, external control integration offers a feasible approach to strengthen interpretations of large-scale prevention trials.
Validation of self-reported family history of myocardial infarction using nationwide health care data.
Wahrenberg A, Leander K, Häbel H … +10 more , Magnusson PKE, Kuja-Halkola R, Bergström G, Lind L, Hagström E, Engström G, Jernberg T, Söderberg S, Östgren CJ, Svensson P
Eur J Epidemiol · 2026 May · PMID 42142221 · Publisher ↗
Self-reported family history of myocardial infarction (MI) is frequently used to assess cardiovascular risk. However, validation of self-reported family history has often been limited to small samples, and to death-certi...
Self-reported family history of myocardial infarction (MI) is frequently used to assess cardiovascular risk. However, validation of self-reported family history has often been limited to small samples, and to death-certificates or self-reported medical history as reference. This study investigated the accuracy of self-reported family history of MI in a population-based cohort, using register data as reference. We included 25,302 participants from the Swedish CArdioPulmonary bioImage Study (SCAPIS) with sufficient register coverage. Self-reported family history of MI in parents and siblings was obtained from the SCAPIS questionnaire. Relatives were identified in the Swedish Multi-Generation Register and occurrences of MI were retrieved from the Swedish National Patient Register and Cause of Death Register. Cohen’s κ, sensitivity, specificity, and predictive values for self-reports were calculated using register-derived diagnoses as reference. The mean age of participants was 57.5 years. The overall agreement of self-reported and register-verified family history was moderate. Reporting family history of MI in any parent or sibling yielded a κ of 0.491, a sensitivity of 57.6% and a specificity of 89.0%. The positive and negative predictive values were 73.9 and 79.6%, respectively. The accuracy of reported maternal and paternal MI were similar, however, reports of early-onset disease showed lower overall κ. Male participants consistently reported family history less accurately than females, and participants with a university degree reported family history more accurately than those without a university education. Compared with register data, self-reported family history of MI underestimates the prevalence of, especially early-onset, MI in close relatives.
Long-term negative divergence in mortality at ages 25-49 years between the United Kingdom and 21 peer countries between 1990 and 2019.
Leon DA, Jdanov D, Medina-Jaudes N … +2 more , Danilova I, Shkolnikov VM
Eur J Epidemiol · 2026 May · PMID 42120865 · Publisher ↗
BACKGROUND: The poor performance of the UK in reducing mortality compared to many other high-income countries following the 2008 financial crisis have been extensively studied, with particular attention to deaths of desp...
BACKGROUND: The poor performance of the UK in reducing mortality compared to many other high-income countries following the 2008 financial crisis have been extensively studied, with particular attention to deaths of despair at working ages. However, longer-term trends in the differences in working-age mortality between the UK and peer countries have not been systematically investigated. METHODS: We compared trends (1990-2019) in age-standardised mortality rates at age 25-49 years in the UK and its constituent parts (England and its 9 standard regions, Wales, Scotland, Northern Ireland) with those of 21 peer countries. FINDINGS: Between 1990 and 2019 the UK went from having relatively low mortality rates at age 25-49 years compared to its peers to having one of the highest. This reflects both the better progress made by many other countries in reducing mortality rates as well as an absolute increase in the UK from 2013. Against the counter-factual that rates in the UK followed the median of the comparator countries (2001-2019) this resulted in 3.1 million excess years of life lost. The divergence in mortality of the UK with its peers was apparent from 1990 and was observed for all constituent parts of the UK and English regions. External cause mortality accounted for much of the divergence in rates between 2001 and 2019 (69% women; 78% men), as did the overlapping categories of drug-related deaths (42%; 28%) and suicides (17%; 20%). Alcohol-related deaths made only a small contribution. INTERPRETATION: The divergence in mortality rates at ages 25-49 years in the UK from peer countries was already apparent from 1990, pre-dating the austerity policies two decades later. Nevertheless, austerity may well have exacerbated this longer-term deterioration in the UKs position. The fact that all areas of the UK showed deterioration relative to peer countries indicates that this is a national problem.
Prevalence, characteristics, and projection of long-term childhood cancer survivors in Sweden.
Extrand E, Heiland EG, Allen G … +3 more , Giertz M, Mogensen H, Brooke HL
Eur J Epidemiol · 2026 May · PMID 42096132 · Publisher ↗
Population-level descriptions of long-term childhood cancer survivors are fundamental to survivorship care and research but seldom available. Accordingly, we aimed to describe long-term childhood cancer survivors at the...
Population-level descriptions of long-term childhood cancer survivors are fundamental to survivorship care and research but seldom available. Accordingly, we aimed to describe long-term childhood cancer survivors at the population-level and project future prevalence. In this register-based study we calculated the absolute number and prevalence proportions of all individuals diagnosed with a childhood cancer (aged 0-14 years, 1958-2018) in Sweden who survived ≥ 5 years post-diagnosis and were alive and residing in Sweden on December 31st, 2023. We also described the clinical and sociodemographic characteristics of this population and presented the observed prevalence over time (1990-2023) and projected prevalence under different mortality assumptions (2024-2040). On December 31st, 2023, there were 8645 long-term childhood cancer survivors in Sweden, equivalent to nearly 1 in 1000 inhabitants (921 persons per million). Leukemias (28.3%) and central nervous system tumors (27.0%) were the most common childhood cancer diagnoses, although the distribution of cancer type varied by attained age. Disease burden in the preceding five years was heterogeneous: approximately 25-30% of the survivors had no recent diagnoses or prescriptions, while a similar proportion experienced substantial morbidity. Most adult survivors were employed (72.0%) and relatively few received sickness benefits (9.7%). From 1990 to 2023, the long-term survivor population tripled in size. Projected mean annual growth was between 1.6% and 2.2%, with the population increasing to approximately 11,400 - 12,600 individuals by 2040. As this heterogeneous population continues growing, our comprehensive description can help plan survivorship care and provide a benchmark for prevalence estimates in settings with less complete data.
Evaluation of clinical utility in emulated clinical trials.
Hruza J, Sjölander A, E Gabriel E … +2 more , Bhatt S, C Sachs M
Eur J Epidemiol · 2026 May · PMID 42081009 · Publisher ↗
Dynamic treatment regimes have been proposed to personalize treatment decisions by utilizing historical patient data, but they may not always improve on the current standard of care. It is thus meaningful to integrate th...
Dynamic treatment regimes have been proposed to personalize treatment decisions by utilizing historical patient data, but they may not always improve on the current standard of care. It is thus meaningful to integrate the standard of care into the evaluation of treatment strategies, and previous works have suggested doing so through the concept of clinical utility. Here we will focus on the comparative component of clinical utility as the average outcome had the full population received treatment based on the proposed dynamic treatment regime in comparison to the full population receiving the "standard" treatment assignment mechanism, such as a physician's choice. Clinical trials to evaluate clinical utility are rarely conducted, and thus, previous works have proposed an emulated clinical trial framework using observational data. However, only one simple estimator was previously suggested, and the practical details of how one would conduct this emulated trial were not detailed. Here, we illuminate these details and propose several estimators of clinical utility based on estimators proposed in the dynamic treatment regime literature. We illustrate the considerations and the estimators in a real data example investigating treatment rules for rheumatoid arthritis, where we highlight that in addition to the standard of care, the current medical guidelines should also be compared to any estimated "optimal" decision rule.
Forecast of dementia prevalence in Germany and subnational regions until 2060 using microsimulation.
Schüssel K, Brückner G, Schröder H … +5 more , Doblhammer G, Jessen F, Ernst J, Caratiola C, Münnich R
Eur J Epidemiol · 2026 May · PMID 42081008 · Publisher ↗
The number of dementia cases is expected to rise in ageing societies with concomitant requirements for healthcare planning. While national figures for countries are widely available, subnational regional heterogeneity is...
The number of dementia cases is expected to rise in ageing societies with concomitant requirements for healthcare planning. While national figures for countries are widely available, subnational regional heterogeneity is rarely considered. This study aimed at projecting national and subnational dementia figures under different scenarios. Dementia and population figures where projected from 2018 to 2060 by microsimulation at the level of 400 counties within Germany. The German MikroSim model was supplemented with dementia-specific data on regional prevalence and national incidence and mortality figures. These data were derived from routine health insurance records of 27 million individuals within the AOK fund during pre-pandemic years 2017 to 2019 and adjusted to the total population of 2018. Scenarios with increasing life expectancy or prevention effects (decreasing dementia incidence by 1 to 2% per year) were calculated. Increasing life expectancy leads to dementia cases rising from 1.4 million in 2018 to 2.1 million in 2060, prevalence from 1.6 to 2.6% and the RWA (ratio of cases per 100 individuals of working age) from 2.6 to 4.7. In prevention scenarios, these figures are substantially lower: between 1.3 and 1.5 million cases, 1.6 and 1.9% prevalence and RWA between 2.9 and 3.4. Existing subnational regional differences in prevalence and RWA will become more pronounced in all scenarios. Growing life expectancy leads to an increase in dementia cases, which may be counteracted by preventive measures. Increasing regional heterogeneity in prevalence and RWA should be considered for health care planning.
Source-specific nitrate and nitrite intake and associations with gastric cancer in the Danish Diet, Cancer and Health Cohort.
Erichsen DW, Kyrø C, Pokharel P … +14 more , Rosthøj S, Bondonno CP, Zhong L, Schullehner J, Sigsgaard T, Dahl C, Hendriksen PF, Dalgaard F, Raaschou-Nielsen O, Hodgson JM, Dahm CC, Olsen A, Tjønneland A, Bondonno NP
Eur J Epidemiol · 2026 May · PMID 42060032 · Publisher ↗
Abstract loading — click title to view on PubMed.
Life expectancy of people with hemophilia in France in 2022.
Tabélé C, Boucekine M, Chambost H … +14 more , Tiade ML, Giraud N, Pradines D, Colle Y, Dalibard V, Kherfellah D, Voltzenlogel R, Beltran Anzola A, Nguyen NAT, Resseguier N, D'Oiron R, Costagliola D, Auquier P, FranceCoag Study group
Eur J Epidemiol · 2026 Apr · PMID 42018064 · Publisher ↗
The life expectancy of adult people with hemophilia has significantly increased thanks to therapeutic advancements, especially after the bloodborne infectious risks were controlled. However, no French data has yet docume...
The life expectancy of adult people with hemophilia has significantly increased thanks to therapeutic advancements, especially after the bloodborne infectious risks were controlled. However, no French data has yet documented this improvement. This study aims to compare the life expectancy of French people with hemophilia to that of the general population. A longitudinal study was conducted on 9,928 people with hemophilia included in the FranceCoag database between 1994 and 2022. Life expectancies were calculated using abridged life tables The life expectancies for people with hemophilia were 66.3 years at age 20 (mild hemophilia = 68.8, moderate = 64.8, severe = 60.4) and 47.5 years at age 40 (mild = 49.2, moderate = 45.9, severe = 42.7). Individuals with mild or moderate hemophilia had higher life expectancies compared to general population (59.8 years at age 20 and 40.7 years at age 40, p < 0.001). Those with severe hemophilia had a similar or slightly higher life expectancy at age 40. The mean differences in LE at ages 20 and 40, compared with the general population, varied markedly by severity, ranging from 0 to 10 years. Taken together, these findings suggest broadly comparable life expectancy in people with hemophilia, although differences remain according to disease severity. This improvement may be attributable both to recent therapeutic innovations and to more frequent and comprehensive medical follow-ups compared to the general population.
Food colouring additives and cancer incidence in the NutriNet-Santé prospective cohort.
Shah S, Hasenböhler A, Javaux G … +17 more , de la Garanderie MP, de Edelenyi FS, Yvroud P, Agaësse C, De Sa A, Huybrechts I, Pierre F, Audebert M, Coumoul X, Julia C, Kesse-Guyot E, Allès B, Hercberg S, Chassaing B, Deschasaux-Tanguy M, Srour B, Touvier M
Eur J Epidemiol · 2026 May · PMID 41954817 · Publisher ↗
Our study aimed to assess potential associations between food colouring additives and cancer incidence in the French NutriNet-Santé cohort. A total of 105,260 adults (78.3% females; mean age 42.0 ± 14.5y) without prevale...
Our study aimed to assess potential associations between food colouring additives and cancer incidence in the French NutriNet-Santé cohort. A total of 105,260 adults (78.3% females; mean age 42.0 ± 14.5y) without prevalent cancer and who completed ≥ 2 24-hour dietary records at baseline were followed for > 7 years. Dietary intakes were assessed using repeated brand-specific 24 h records, and cumulative time-dependent exposure to food additives was evaluated through multiple composition databases and ad-hoc laboratory assays in food matrices. Associations between exposures to food colouring additives (sex-specific tertiles if proportion of exposed participants > 2/3, or non-exposed/lower/higher exposed based on sex-specific median otherwise) and cancer incidence were assessed using multivariable Cox models. We identified 4,226 incident cancer cases (508 prostate, 1,208 breast [387 premenopausal and 821 postmenopausal], and 352 colorectal). Total food colouring additives were associated with higher overall [HR (95%CI): 1.14 (1.05-1.24); absolute risk at age 60: 13.3% (higher consumers) vs. 12.1% (lower/non-consumers)], breast [1.21 (1.03-1.42); 5.7%, 4.8%], and postmenopausal breast [1.32 (1.09-1.61); 14.9%, 12.5%] cancer incidence. After False Discovery Rate correction, only plain caramel (European code: E150a) was associated with overall cancer [1.15 (1.07-1.25); 14.0%, 12.1%] and beta-carotene (E160a) with overall [1.16 (1.07-1.25); 13.7%, 11.9%] and breast cancer [1.41 (1.23-1.62); 6.2%, 4.4%]. This study provides novel exploratory evidence linking colouring additives to cancer incidence. Further research is needed to elucidate underlying mechanisms. Findings support recommendations to limit exposure to non-essential food additives (i.e. used primarily for sensory or aesthetic purposes rather than for food safety or preservation) and inform regulatory reassessment.
CohortCharacteristics: an R package for population characterisation in observational studies using the OMOP common data model.
Du M, Prats-Uribe A, Mercadé-Besora N … +23 more , Lopez-Guell K, Guo Y, Alcalde-Herraiz M, Chen X, Delmestri A, Man WY, Duarte-Salles T, Palomar A, Giuliodori A, Brađašević E, Jezidžić A, Bräuner E, Bruun S, Verhamme K, Mosseveld M, Brash JT, Vojinovic D, Kaczmarczyk I, Mendez A, Rijnbeek P, Prieto-Alhambra D, Burn E, Català M
Eur J Epidemiol · 2026 Apr · PMID 41933108 · Full text
Describing cohort characterisation ensures comparability and reproducibility in multi-database observational studies. To address this need, we developed CohortCharacteristics, an open-source R package that facilitates st...
Describing cohort characterisation ensures comparability and reproducibility in multi-database observational studies. To address this need, we developed CohortCharacteristics, an open-source R package that facilitates standardised cohort characterisation in datasets mapped to the Observational Medical Outcomes Partnership (OMOP) Common Data Model (CDM). This study aims to explain the development of the package and demonstrate its core functionality. We developed CohortCharacteristics, an open-source R package that can perform cohort characterisation for various types of databases. To demonstrate its functionality, we then used CohortCharacteristics to generate descriptive statistics on demographics, comorbidities, medication exposures, cohort overlap, and timing of cohort entries. The study included data from CPRD GOLD (UK), DK-DHR (Denmark), IPCI (Netherlands), IQVIA Longitudinal Patient Database Belgium (IQVIA LPD Belgium), IQVIA DA Germany, NAJS (Croatia), and SIDIAP (Spain), all mapped to the OMOP CDM. The CohortCharacteristics R package is freely available on CRAN with detailed vignettes and documentation on its functionality. Cohort characteristics were generally consistent across databases, with similar age distributions and female representation. CPRD GOLD, NAJS, and SIDIAP exhibited higher prescribing rates for respiratory, cardiovascular, and nervous system medications, while IQVIA databases and DK-DHR reported lower rates. Timing analysis showed that dementia diagnoses typically followed insomnia diagnoses in several databases, supporting existing literature. Antipsychotic prescriptions often occurred after dementia diagnosis, reflecting prescribing practices aligned with clinical guidelines. CohortCharacteristics enables consistent cohort characterisation across a network of data mapped to the OMOP CDM, thereby improving transparency in multi-database research. The package's functionality, demonstrated in this study, illustrates its applicability in observational studies with OMOP CDM data.
Vegetarian and vegan diets and cancer incidence: a systematic review and meta-analysis of prospective studies.
Aune D, Schlesinger S, Sobiecki JG
Eur J Epidemiol · 2026 May · PMID 41879966 · Publisher ↗
Several studies have suggested that vegetarian and vegan vs. non-vegetarian diets are associated with lower cancer risk overall, however, results for specific cancer sites have been less consistent. We conducted a system...
Several studies have suggested that vegetarian and vegan vs. non-vegetarian diets are associated with lower cancer risk overall, however, results for specific cancer sites have been less consistent. We conducted a systematic review and meta-analysis of prospective studies on vegetarian and vegan diets and cancer incidence to clarify the associations across cancer sites. PubMed and Embase databases were searched for relevant studies up to 5 July 2025. Summary relative risks (RRs) and 95% confidence intervals (95% CIs) were calculated for the association between vegetarian and vegan diets and cancer incidence. World Cancer Research Fund (WCRF) criteria was used to evaluate the strength of the evidence. Seventeen publications (seven prospective studies) were included. The summary RRs (95% CIs) for vegetarians vs. non-vegetarians was 0.87 (0.84-0.91, I = 0%, n = 4 studies) for total cancer incidence, 0.55 (0.36-0.86, I = 32%, n = 4) for stomach cancer, 0.86 (0.76-0.97, I = 14%, n = 6) for colorectal cancer, 0.79 (0.67-0.93, I = 38%, n = 7) for colon cancer, 0.55 (0.31-0.97, I = 0%, n = 2) for proximal colon cancer, 0.77 (0.62-0.95, I = 0%, n = 5) for pancreatic cancer, 0.79 (0.66-0.94, I = 0%, n = 4) for melanoma, 0.92 (0.86-0.99, I = 0%, n = 7) for breast cancer, 0.81 (0.69-0.95, I = 0%, n = 3) for postmenopausal breast cancer, 0.78 (0.62-0.98, I = 0%, n = 5) for bladder cancer, and 0.76 (0.63-0.93, I = 0%, n = 4) for non-Hodgkin's lymphoma. In addition, non-statistically significant inverse associations were observed for some cancers, with summary RRs of 0.85 (0.70-1.04, I = 0%, n = 6) for lung cancer, 0.83 (0.68-1.02, I = 0%, n = 5) for ovarian cancer, and 0.87 (0.75-1.00, I = 43%, n = 6) for prostate cancer. Results for other cancer sites were imprecise or near the null. The summary RRs (95% CIs) for vegans vs. non-vegetarians were 0.77 (0.70-0.85, I = 0%, n = 3) for total cancer, 1.02 (0.71-1.48, I = 42%, n = 3) for colorectal cancer, 0.80 (0.64-1.00, I = 0%, n = 4) for breast cancer, and 0.87 (0.50-1.49, I = 49%, n = 3) for prostate cancer. BMI explained a moderate part of the associations. The strength of evidence [judging the likelihood of causality] for vegetarian diets and total, colorectal, colon and breast cancer was judged as probable, and limited suggestive for stomach, pancreatic, and bladder cancers, melanoma and non-Hodgkin's lymphoma, and for vegan diets and total cancer and breast cancer was considered limited-suggestive. Vegetarian diets compared to non-vegetarian diets are associated with reduced risk of total cancer and seven specific cancer types, while vegan diets are associated with reduced risk of total and breast cancer. Although further studies are needed to assess the long-term adherence to vegetarian and vegan diets and cancer incidence and across less investigated cancers, these results support recommendations to adopt much more plant-based diets for cancer prevention.
Impact of hypothyroidism and hyperthyroidism on endometrial cancer incidence: results from a large population-based cohort study.
Jensen A, Nøhr B, Christensen J … +4 more , Gottschau M, Lindquist S, Pedersen J, Kjaer SK
Eur J Epidemiol · 2026 May · PMID 41879965 · Publisher ↗
The limited number of studies examining the association between thyroid diseases and endometrial cancer have yielded inconsistent findings. The study aimed to examine the association between hypothyroidism and hyperthyro...
The limited number of studies examining the association between thyroid diseases and endometrial cancer have yielded inconsistent findings. The study aimed to examine the association between hypothyroidism and hyperthyroidism and the risk of endometrial cancer using comprehensive nationwide register data from Denmark. We conducted a population-based cohort study including 1,057,937 women born in Denmark between 1960 and 1997. Information on thyroid disease diagnoses, cancer diagnoses, covariates, migration, and vital status was obtained from nationwide Danish health and administrative registers. Cox proportional hazards models were used to estimate adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for endometrial cancer overall and for type I tumors. A landmark analysis examined risks associated with exposures before age 40, and pseudo-observation methods estimated absolute risk differences. During a median follow-up of 17.5 years, 1,159 women were diagnosed with endometrial cancer. Women with hypothyroidism had a higher rate of overall endometrial cancer (HR: 1.53, 95% CI: 1.22-1.93) and type I tumors (HR: 1.64, 95% CI: 1.12-2.41). These associations were consistent across subgroups defined by menopausal status and time since diagnosis. No association was observed for hyperthyroidism (HR: 1.14, 95% CI: 0.80-1.62). In the landmark analysis, hypothyroidism remained associated with an increased endometrial cancer rate, but the absolute risk difference by age 60 was modest and not statistically significant. In conclusion, hypothyroidism was associated with a modestly increased rate of endometrial cancer, while no association was observed for hyperthyroidism. These findings support further investigation into thyroid function and endometrial carcinogenesis.
A study evaluation framework for measuring cognition: lessons learned in cross-national contexts from four English-speaking aging cohorts.
Hayat S, Assaad S, Ahmed N … +2 more , Brayne C, Steptoe A
Eur J Epidemiol · 2026 May · PMID 41874903 · Publisher ↗
The Harmonized Cognitive Assessment Protocol (HCAP) is a detailed battery assessing cognition among older people used by studies across the world. Data harmonization is a key priority for HCAP studies. Errors or artefact...
The Harmonized Cognitive Assessment Protocol (HCAP) is a detailed battery assessing cognition among older people used by studies across the world. Data harmonization is a key priority for HCAP studies. Errors or artefacts introduced during data collection become embedded in the dataset and cannot be fully resolved through post-hoc statistical adjustment alone. We used a mixed-methods approach using established theories from the existing literature on methodologies of longitudinal studies and the implementation of HCAP in four English-speaking studies adopting the same protocol. Through a detailed investigation involving the English Longitudinal Study of Ageing (ELSA), the Health and Retirement Study (HRS), The Irish Longitudinal Study on Ageing (TILDA), and the Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA), we identified 60 factors contributing to the development of a conceptual framework for the evaluation and implementation of HCAP. We present this framework and a prototype checklist as a tool for providing a transparent and structured approach to improve data quality, cross-country comparability and for identifying, mitigating, and monitoring sources of bias. The framework consisting of four broad headings: (1) Organisation and design, (2) Competency of personnel and systems, (3) Implementation and outputs, and (4) Feedback and communication. This framework aims improves data quality at the point of collection, designed to complement and not replace, post-hoc statistical harmonization. By strengthening data at the outset enables subsequent harmonization to be more robust. We recommend studies seeking cross-national comparability to give careful consideration to operational aspects of fieldwork.
A target trial emulation of the CONFIRM study with an extension to subgroups: an example for relapsing-remitting multiple sclerosis.
Verweij S, Bijlsma MJ, Rengerink KO … +7 more , Hillert J, Forsberg L, Mouresan EF, Glaser A, Fox RJ, Hak E, Mol P
Eur J Epidemiol · 2026 Mar · PMID 41849089 · Publisher ↗
Conclusions of target trial emulations (TTEs) are often limited to trial-like populations, leaving knowledge gaps for neglected subpopulations. We propose a two-step TTE approach to create methodologically-justified evid...
Conclusions of target trial emulations (TTEs) are often limited to trial-like populations, leaving knowledge gaps for neglected subpopulations. We propose a two-step TTE approach to create methodologically-justified evidence for these subpopulations. Using the CONFIRM trial as an illustrative example, we first emulated the trial using strict eligibility criteria, and extended the analysis to a pragmatic scenario including trial-ineligible subgroups, allowing comparison of treatment effectiveness between these two study populations. CONFIRM studied efficacy of dimethyl fumarate (DMF) and glatiramer acetate (GA) in reducing relapses in 709 relapsing-remitting multiple sclerosis (RRMS) patients. A retrospective cohort study was designed using the Swedish multiple sclerosis registry. Outcomes were the annualized relapse rate (ARR) and the probability of a relapse, adjusted for covariates and overlap weights. Using strict eligibility criteria, 288 RRMS patients were included, whereas the pragmatic scenario comprised 1831 patients. The ARR ratio for DMF versus GA was 0.78 (95%CI 0.59–1.05) in CONFIRM, 0.75 (95%CI: 0.44–1.29) in the strict scenario, and 0.58 (95% CI: 0.36–0.94) in the pragmatic scenario. Similarly, the hazard ratio was 0.92 (95%CI 0.70–1.22), 1.05 (95%CI: 0.83–1.33) and 0.56 (95%CI: 0.46–0.67), respectively. Post-hoc analysis implied existence of calendar bias. Comparable results between CONFIRM and the strict scenario were observed, while the second step demonstrates how methodological advantages of TTEs can be extended to excluded subgroups. This study can be used as guiding example for future observational studies that wish to incorporate the TTE framework, as well as embrace the wider variety in the patient population often ignored in experimental settings.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe