European Journal of Epidemiology最新文献

Previous studies have investigated the role of metabolic factors in risk of hematological malignancies with contradicting findings. Existing studies are generally limited by potential concern of reverse causality and confounding by inflammation. Therefore, we aimed to investigate the associations of glucose, lipid, and apolipoprotein biomarkers with the risk of hematological malignancy. We performed a study of over 560,000 individuals of the Swedish AMORIS cohort, with measurements of biomarkers for carbohydrate, lipid, and apolipoprotein metabolism during 1985–1996 and follow-up until 2020. We conducted a prospective cohort study and used Cox models to investigate the association of nine different metabolic biomarkers (glucose, total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), LDL-C/HDL-C, triglyceride (TG), apolipoprotein B (ApoB), apolipoprotein A-I (ApoA I), and ApoB/ApoA-I) with risk of hematological malignancy, after excluding the first five years of follow-up and adjustment for inflammatory biomarkers. We observed a decreased risk of hematological malignancy associated with one SD increase of TC (HR 0.93; 95% CI 0.91–0.96), LDL-C (HR 0.94; 95% CI 0.91–0.97), HDL-C (HR 0.92; 95% CI 0.86–0.99), and ApoA-I (HR 0.96; 95% CI 0.93–0.996). Our study highlights a decreased risk of hematological malignancy associated with a higher level of TC, LDL-C, HDL-C, and ApoA-I.

Background

Accumulating evidence suggest that Epstein-Barr virus (EBV) is crucial in the development of multiple sclerosis (MS), with inadequate infection control possibly contributing to disease onset. Past infectious mononucleosis (IM) has been found to interact with smoking, obesity, and sun exposure. We aimed to investigate potential interactions between a history of IM and the following risk factors for MS: passive smoking, alcohol consumption, fish consumption, vitamin D status, adolescent sleep duration and sleep quality.

Methods

We analyzed data from a Swedish population-based case-control study (3128 cases and 5986 controls). Subjects were categorized based on IM status and each exposure variable and compared regarding MS risk by calculating odds ratios (OR) with 95% confidence intervals (CI) using logistic regression models. Additive interaction between aspects of IM status and each exposure was assessed by calculating the attributable proportion due to interaction (AP) with 95% CI.

Results

The OR of developing MS among those who reported a history of IM was 1.86 (95% CI 1.63–2.12), compared with those who had not suffered from IM. We observed synergistic effects between a history of IM and each exposure variable with respect to risk of MS, with significant APs ranging between 0.20 and 0.35.

Conclusions

The concept of EBV infection as a crucial factor for MS gains further support from our findings suggesting that MS risk factors synergize with a history of IM in disease development. Targeting modifiable MS risk factors that impede effective immune regulation of the virus holds promise for preventive interventions.

Effect heterogeneity analyses using causal machine learning algorithms have gained popularity in recent years. However, the interpretation of estimated individualized effects requires caution because insights from these data-driven approaches might be misaligned with the contextual needs of a human audience. Thus, a practical framework that integrates advanced machine learning methods and decision-making remains critically needed to achieve effective implementation and scientific communication. We introduce a 2-step framework to identify characteristics associated with substantial effect heterogeneity in a practically relevant format. The proposed framework applies distinct sets of covariates for (i) estimation of individualized effects and (ii) subgroup discovery and shows the subgroups with heterogeneity based on highly interpretable if-then rules. By referring to existing metrics of interpretability, we describe how each step contributes to leveraging a theoretical advantage of machine learning models while creating an interpretable and practically relevant framework. We applied the pragmatic subgroup discovery framework for the Look AHEAD (Action for Health in Diabetes) trial to assess practically relevant and comprehensive insights into the effect heterogeneities of intense lifestyle intervention for individuals with diabetes on cardiovascular mortality. Our analysis identified (i) individuals with history of cardiovascular disease and myocardial infarction had the least benefit from the intervention, while (ii) individuals with no history of cardiovascular diseases and HbA1c < 7% received the highest benefit. In summary, our practical framework for heterogeneous effects discovery could be a generic strategy to ensure both effective implementation and scientific communication when applying machine learning algorithms in epidemiological research.

Despite the growing population of young cancer survivors of reproductive age, the risk of cancer in offspring born to female cancer survivors has yielded inconsistent results. Therefore, this study aimed to investigate the risk of cancer among the offspring of female cancer survivors by maternal age at delivery, maternal age at cancer diagnosis, maternal cancer type, and the time interval between cancer diagnosis and pregnancy. Using nationwide retrospective mother–child linked data from the Korean National Health Insurance Service, we included the first child (N = 8031) of female cancer survivors aged < 40 years after excluding thyroid cancer survivors and matched controls (N = 24,093) between 2005 and 2019. Subgroup analysis was performed according to maternal age at delivery, maternal age at cancer diagnosis, maternal cancer type, and the interval between cancer diagnosis and delivery. Among the offspring, 19 children of cancer survivors and 30 in the control group were diagnosed with cancer, with a mean age of 2.0 years at diagnosis. The most prevalent cancer type was leukemia (26.5%), followed by liver tumor (10.2%) and brain tumor (8.2%). The hazard ratio (HR) for cancer in the offspring of female cancer survivors was 1.91 (95% confidence interval (CI) = 1.07–3.38) demonstrating consistently high risk over the follow-up period. HRs for cancer risk in offspring were high across all subgroups despite the low statistical power. Our study indicated that offspring born to maternal cancer survivors had an increased risk of cancer.

Recent advancements in machine learning (ML) for analyzing heterogeneous treatment effects (HTE) are gaining prominence within the medical and epidemiological communities, offering potential breakthroughs in the realm of precision medicine by enabling the prediction of individual responses to treatments. This paper introduces the methodological frameworks used to study HTEs, particularly based on a single randomized controlled trial (RCT). We focus on methods to estimate conditional average treatment effect (CATE) for multiple covariates, aiming to predict individualized treatment effects. We explore a range of methodologies from basic frameworks like the T-learner, S-learner, and Causal Forest, to more advanced ones such as the DR-learner and R-learner, as well as cross-validation for CATE estimation to enhance statistical efficiency by estimating CATE for all RCT participants. We also provide a practical application of these approaches using the Preventing Overweight Using Novel Dietary Strategies (POUNDS Lost) trial, which compared the effects of high versus low-fat diet interventions on 2-year weight changes. We compared different sets of covariates for CATE estimation, showing that the DR- and R-learners are useful for the estimation of CATE in high-dimensional settings. This paper aims to explain the theoretical underpinnings and methodological nuances of ML-based HTE analysis without relying on technical jargon, making these concepts more accessible to the clinical and epidemiological research communities.

Night work and circadian disruption are linked to major public health challenges, e.g. cancer, cardiometabolic disease, and accidents. We established the 1001 nights-cohort to explore mechanisms underlying health effects of night work and circadian disruption. 1075 female hospital employees participated from September 2022 to April 2024. The data collection included a questionnaire, a blood sample, anthropometric measures, and sleep actigraphy and sleep diaries across 14 days. In subsamples, light exposure, physical activity, skin temperature, and blood glucose were measured continuously for 7 days, and saliva samples were collected five times across one day. The cohort consists of 2- and 3-shift workers with night work (66%), permanent night workers (7%), permanent evening workers or 2-shift workers without night work (9%), and permanent day workers (18%). Data comprise 4553 day shifts, 997 evening shifts, 1963 night shifts, and 6458 days without work. The poorest health was observed among permanent night workers and the group of shift workers without night work. The 1001 nights-cohort is the most comprehensive data within night work and working hour research due to the combination of questionnaires, biomarkers, technical measurements, and possibilities for linkage to historical and future register-based information about working hours from the Danish Working Hour Database (DAD) and diagnoses. With its repeated measurements within the same individual, the cohort will advance research on physiological and behavioral mechanisms underlying health effects of working hours, night work, and circadian disruption and deliver important scientific input for updating guidelines on healthy scheduling of working hours.

BiCoVac is a population-based Danish cohort aiming to examine whether Coronavirus disease 2019 (COVID-19) vaccines are associated with non-specific symptoms beyond the specific protection of COVID-19. Data were collected by four questionnaire surveys between May 2021 and July 2022 and the questionnaire distribution was aligned with the Danish COVID-19 vaccination program. All surveys collected self-reported information on symptoms (e.g., headache, nausea, and fatigue). The baseline survey additionally gathered information on lifestyle and health. Survey data were combined with data from the Danish registers including information on COVID-19 vaccination and COVID-19 test results. A total of 911,613 (25% of all Danish citizens aged 16 to 65) were randomly sampled for the cohort and 252,401 initiated the baseline questionnaire. Of these, 59% (n = 149,070) participated in the 1st follow-up, 43% (n = 107,655) in the 2nd follow-up, and 25% (n = 63,737) in the 3rd follow-up. Women and individuals above 40 years of age were more likely to participate. Among vaccinated respondents, 25–38% reported moderate to severe immediate symptoms following COVID-19 vaccination, varying by vaccine doses. Females, younger individuals, and those with prior COVID-19 reported more immediate symptoms. Results of potential non-specific symptoms following COVID-19 vaccination did not reveal higher risk of involuntary movements among vaccinated individuals compared to unvaccinated individuals. Currently (December 2024), we are further investigating the effects of COVID-19 vaccines on other non-specific symptoms and exploring whether specific characteristics render some individuals more susceptible to report non-specific symptoms. In addition, long-term symptoms following COVID-19 are being investigated.

Background

Previous studies have reported lower net survival probabilities for socioeconomically deprived patients, using non-deprivation specific lifetables. Not accounting for the social gradient in background mortality could potentially overestimate the effect of deprivation on net survival. The aim of this study was to estimate the impact of taking into account the social gradient of expected mortality in the general population on the study of the social gradient of survival of people with cancer.

Methods

French cancer registry data was analyzed, with 190,902 incident cases of nineteen cancer sites between 2013 and 2015. Deprivation was measured using the European deprivation index (EDI). Net survival was estimated thanks to additive models with French lifetables stratified on deprivation level with the EDI, using the non-parametric Pohar-perme method and flexible excess hazard modelling with multidimensional penalized splines, firstly with non-specific lifetables then with the deprivation specific-lifetables.

Results

A significant effect of EDI on excess mortality hazard (EMH) remained when using the deprivation-specific lifetables for colorectal, lung cancer and melanoma in both sexes, and esophagus, bladder, head and neck and liver cancer for men, and breast, cervix and uterine cancer for women. The only site where the effect of EDI on EMH was no longer significant when using deprivation-specific lifetables was prostate cancer.

Conclusions

The use of deprivation-specific lifetables confirms the existence of a social gradient in cancer survival, indicating that these inequalities do not result from inequalities in background mortality. Development of such deprivation-specific lifetables for future years is crucial to understand mechanisms of social inequalities and work towards reducing the social burden.

The risk over a given time span can be calculated as one minus the exponentiated value of the negative of the integral of the incidence density function (or hazard rate function) over that time span. This relationship is widely used but, in the few instances where textbooks have presented it, the derivations of it tend to be purely mathematical. I first review the historical contexts, definitions, distinctions and links. I then offer a more intuitive heuristic approach that draws on the conceptualization of a person-year in Edmonds’ 1832 definition of the force of mortality, and on the number of replacements in a dynamic population. Similarly I show how the Nelson-Aalen risk estimator can be seen in the context of this historical conceptualization of a person-year, scaled to the experience of a dynamic population of (constant) size 1.