European Journal of Epidemiology最新文献

The International Nuclear Workers Study (INWORKS) contributes knowledge on the dose-response association between predominantly low dose, low dose rate occupational exposures to penetrating forms of ionizing radiation and cause-specific mortality. By extending follow-up of 309,932 radiation workers from France (1968–2014), the United Kingdom (1955–2012), and the United States (1944–2016) we increased support for analyses of temporal variation in radiation-cancer mortality associations. Here, we examine whether age at exposure, time since exposure, or attained age separately modify associations between radiation and mortality from all solid cancers, solid cancers excluding lung cancer, lung cancer, and lymphohematopoietic cancers. Multivariable Poisson regression was used to fit general relative rate models that describe modification of the linear excess relative rate per unit organ absorbed dose. Given indication of greater risk per unit dose for solid cancer mortality among workers hired in more recent calendar years, sensitivity analyses considering the impact of year of hire on results were performed. Findings were reasonably compatible with those from previous pooled and country-specific analyses within INWORKS showing temporal patterns of effect measure modification that varied among cancers, with evidence of persistent radiation-associated excess cancer risk decades after exposure, although statistically significant temporal modification of the radiation effect was not observed. Analyses stratified by hire period (< 1958, 1958+) showed temporal patterns that varied; however, these analyses did not suggest that this was due to differences in distribution of these effect measure modifiers by hire year.

Twins suffer a disproportionately higher burden of adverse perinatal outcomes than singletons. However, the degree to which preterm delivery shapes the relationship between abruption and perinatal mortality in twins is unknown. Through causal mediation decomposition, we examine how preterm delivery mediates the effect of abruption on perinatal mortality among twins using the US-matched multiple birth data (1995–2000). We estimated the hazard ratio (HR) from Cox models with gestational age as the timescale. We decomposed the total effect (TE) into counterfactual natural direct (NDE) and natural indirect (NIE) effects. 557,220 matched twin births, 1.3% (n = 7032) resulted in abruption with higher perinatal mortality rates than non-abruption births (143 versus 36 per 1000 births, respectively) and a 4.53-fold (95% confidence interval [CI]: 4.23, 4.82) increased hazard of perinatal mortality. HRs for NDE and NIE were 3.05 (95% CI: 2.84, 3.24) and 1.49 (95% CI: 1.49, 1.47, 1.50), respectively, and the proportion mediated (PM) was 41%. PM increased as the gestational age at delivery decreased. Associations persisted after correction for unmeasured confounders. The best strategies to improve perinatal delivery are delivery when abruption complicates twin pregnancies at term gestations and expectant management (avoiding early preterm delivery), if feasible, when abruption complicates twin pregnancies at preterm gestations.

Plasma proteomics could enhance risk prediction for multiple diseases beyond conventional risk factors or polygenic scores (PS). To assess utility of proteomics for risk prediction of ischemic heart disease (IHD) compared with conventional risk factors and PS in Chinese and European populations. A nested case-cohort study measured plasma levels of 2923 proteins using Olink Explore panel in ~ 4000 Chinese adults (1976 incident IHD cases and 2001 sub-cohort controls). We used conventional and machine learning (Boruta) methods to develop proteomics-based prediction models of IHD, with discrimination assessed using area under the curve (AUC), C-statistics and net reclassification index (NRI). These were compared with conventional risk factors and PS in Chinese and in 37,187 Europeans. Overall, 446 proteins were associated with IHD (false discovery rate < 0.05) in Chinese after adjustment for conventional cardiovascular disease risk factors. Proteomic risk models alone yielded higher C-statistics for IHD than conventional risk factors or PS (0.855 [95%CI 0.841–0.868] vs. 0.845 [0.829–0.860] vs 0.553 [0.528–0.578], respectively). Addition of 446 proteins to PS improved C-statistics to 0.857 (0.843–0.871) and NRI by 109.1%; and addition to conventional risk factors improved C-statistics to 0.868 (0.854–0.882) and NRI by 86.9%. Boruta analysis identified 30 proteins accounting for ~ 90% of improvement in NRI for IHD conferred by all 2923 proteins. Similar proteomic panels yielded comparable improvements in risk prediction of IHD in Europeans. Plasma proteomics improved risk prediction of IHD beyond conventional risk factors and PS and could enhance precision medicine approaches for primary prevention of IHD.

Healthcare workers may have different risk for severe outcomes compared with the general population during diverse crises. This paper introduces the concept of healthcare worker versus population hazard (HPH), the risk of an outcome of interest in active healthcare workers compared with the general population they serve. HPH can be expressed with relative risk (HPH(r)) and absolute risk difference (HPH(a)) metrics. Illustrative examples are drawn from infectious outbreaks, war, and the COVID-19 pandemic on death outcomes. HPH can be extreme for lethal outbreaks (HPH(r) = 30 to 143, HPH(a) = 8 to 91 per 1000 for Ebola deaths in 3 Western African countries in 2013-5), and modestly high in relative terms and very high in absolute terms for protracted, major armed conflicts (HPH(r) = 1.38 and HPH(a) = 10.2 for Syria during 2011–2024). Conversely, healthcare workers had 8-12-fold lower risk than the population they served for pandemic excess deaths (physicians in USA) or COVID-19 deaths (physicians in Ontario, healthcare workers in Finland), while healthcare workers in Indonesia did not have this advantage for COVID-19 deaths versus the general population. HPH is susceptible to data inaccuracies in numbers of at-risk populations and of outcomes of interest. Importantly, inferences about healthcare worker risk can be misleading, if deaths of retired healthcare workers contaminate the risk calculations– as in the case of misleading early perceptions of exaggerated COVID-19 risk for healthcare professionals. HPH can offer useful insights for risk assessment to healthcare professionals, the general public, and policy makers and may be useful to monitor for planning and interventions during crises.

The Jiangsu Biobank for the Prevention and Control of Diabetes (JBPCD) is a community-based prospective cohort study initiated in Jiangsu province, to examine the burden of complications associated with type 2 diabetes mellitus (T2DM). This study aims to better understand the incidence, risk factors, and long-term outcomes of T2DM complications to inform prevention and control strategies. From October 2013 to July 2014, 20,053 T2DM patients (7,862 males and 12,191 females) were recruited from Suzhou city and Huai’an city. Baseline data were collected through questionnaire survey, physical examination and biochemical testing, with blood samples stored in a biobank. The follow-up focused on the incidence and mortality related to T2DM complications, linked to national and local medical datasets. Between August 2019 and October 2020, the repeated assessments were completed for 13,973 participants, including questionnaire, physical examination and repeated blood collection. The study identified 1,479 new cancer cases and 3,324 cardiovascular disease cases, with an overall mortality rate of 25.66 per 1,000 person-years. JBPCD welcomes research collaborations and data access requests via email. Currently, there are no plans to provide cohort data for free public access, but specific proposals for further collaboration are welcome. For further information and collaboration, please email [jswuming@vip.sina.com] and [sc@njmu.edu.cn].

Non-response to surveys is a common problem; even more so during the COVID-19 pandemic with social distancing measures challenging data collection. As respondents often differ from non-respondents, this can introduce bias. The goal of the current study was to see if we can reduce bias and restore sample representativeness in a series of COVID-19 surveys embedded within five UK cohort studies by using the rich data available from previous waves of data collection. Three surveys were conducted during the pandemic across five UK cohorts: National Survey of Health and Development (NSHD, born 1946), 1958 National Child Development Study (NCDS), 1970 British Cohort Study (BCS70), Next Steps (born 1989-90) and Millennium Cohort Study (MCS, born 2000-02). Response rates in the COVID-19 surveys were lower compared to previous waves, especially in the younger cohorts. We identified bias due to systematic non-response in several variables, with more respondents in the most advantaged social class and among those with higher childhood cognitive ability. Making use of the rich data available pre-pandemic in these longitudinal studies, the application of non-response weights and multiple imputation was successful in reducing bias in parental social class and childhood cognitive ability, nearly eliminating it for the former. Surveys embedded within existing cohort studies offer a clear advantage over cross-sectional samples collected during the pandemic in terms of their ability to mitigate selection bias. This will enhance the quality and reliability of future research studying the medium and long-term effects of the pandemic.

The China Prospective Multi-Center Birth Cohort Study was launched in 2022. In collaboration with medical centers in 12 cities, it aims to establish a high-quality, multidimensional cohort comprising 20,000 natural pregnancy and assisted reproductive families. As of June 26, 2024, 12,911 pregnant women have participated in this study, and 161,122 biological samples have been collected. These samples cover four critical periods (early pregnancy, mid-pregnancy, late pregnancy, and postpartum) and comprise 10 different types such as serum, plasma, and urine. The study has collected comprehensive information from early pregnancy to newborns. The participants have an average age of 29.76 years, an average height of 160.46 cm, an average pre-pregnancy BMI of 23.11, and an average BMI of 27.25 before delivery. The cohort includes individuals from 26 ethnic groups, with 25 minority groups comprising 5.03% of the population. Guizhou Province exhibits the highest percentage of ethnic minorities at 24.96% and Guangdong Province owns the highest proportion (12.22%) of women with two or more children. The prevalence of thalassemia in Guangdong Province is seven times higher than in other provinces. Among pregnant women over 35 years old, the prevalence of gestational diabetes mellitus is twice that of women under 35. Additionally, the prevalence of preeclampsia in women with assisted reproductive pregnancies is more than twice that of those with natural pregnancies. The study extensively collected diverse data and biological samples, making this cohort an ideal candidate for DOHaD field and multi-omics research.

In causal inference, parametric models are usually employed to address causal questions estimating the effect of interest. However, parametric models rely on the correct model specification assumption that, if not met, leads to biased effect estimates. Correct model specification is challenging, especially in high-dimensional settings. Incorporating Machine Learning (ML) into causal analyses may reduce the bias arising from model misspecification, since ML methods do not require the specification of a functional form of the relationship between variables. However, when ML predictions are directly plugged in a predefined formula of the effect of interest, there is the risk of introducing a “plug-in bias” in the effect measure. To overcome this problem and to achieve useful asymptotic properties, new estimators that combine the predictive potential of ML and the ability of traditional statistical methods to make inference about population parameters have been proposed. For epidemiologists interested in taking advantage of ML for causal inference investigations, we provide an overview of three estimators that represent the current state-of-art, namely Targeted Maximum Likelihood Estimation (TMLE), Augmented Inverse Probability Weighting (AIPW) and Double/Debiased Machine Learning (DML).

Background: Dietary strategies for type 2 diabetes (T2DM) prevention have mainly focused on solid foods and nutrients. Emanating evidence suggests that beverage consumption in adulthood may also influence T2DM development, whereas the role of beverages during adolescence remains unknow.

Objective: To examine adolescent beverages consumption, and their changes from adolescence to adulthood in relation to T2DM risk in adulthood.

Methods: This prospective cohort study, conducted within the Nurses' Health Study II (NHS II), enrolled 41,317 women who completed a food-frequency questionnaire (FFQ) regarding their diet in high school and had no diabetes, cardiovascular disease, or cancer at baseline (1997). Beverage consumption including coffee, tea, regular or diet soda, fruit juice or milk, was assessed using the FFQ. Cox proportional hazards models were utilized to estimate hazard ratios (HRs) for the association between beverage consumption in adolescence and risk of incident type 2 diabetes (T2DM) in adulthood, adjusting for potential confounders.

Results: During 725,650 person-years of follow-up, 2,844 participants developed T2DM. After adjustment for demographic, lifestyle and dietary risk factors, comparing ≥ 1 serving/day with non-consumers, adolescent coffee [HR, 0.86 (95% confidence interval: 0.75 to 0.98); P-trend = 0.02)] and orange juice [HR, 0.83 (0.71 to 0.96); P-trend = 0.0008)] consumption was associated with lower T2DM risk, whereas, regular soda [HR, 1.37 (1.20 to 1.57); P-trend < 0.0001)] and iced tea [HR, 1.41 (1.21 to 1.65); P-trend < 0.0001)] intake was associated with higher T2DM risk. Increased coffee intake from adolescence to adulthood in 1991 was associated with a lower T2DM risk [HR, 0.70 (0.61 to 0.80); P-trend < 0.0001), comparing ≥ + 3 servings/day with no change], whereas the opposite was observed for increased regular soda [HR, 1.20 (1.06 to 1.35); P-trend = 0.004), comparing ≥ + 1 or more servings/week with no change)] and diet soda consumption [HR, 1.59 (1.41 to 1.80); P-trend = 0.0002), comparing ≥ + 2 servings/day with no change].

Conclusion: Adolescent consumption of coffee or orange juice intake was associated with a lower risk of T2DM, whereas the opposite was observed for intake of regular soda or iced tea. In addition, increased coffee intake was associated with a lower diabetes risk, whereas the opposite was observed for regular or diet soda intake. These data highlight a potentially important role of beverage intake at early life in the etiology of diabetes during adulthood.

This study evaluated the relationship between Covid-19 vaccination and menstrual bleeding disturbances using a large national registry linkage including 666,467 women between 20 and 40 years of age residing in Norway on January 1st, 2019. Information on vaccination-BNT162b2 and mRNA-1273 - was obtained from the Norwegian vaccination registry. Diagnoses of menstrual disturbances (absent/scanty, excessive, irregular/frequent menstruation, and intermenstrual bleeding) was obtained from the general practitioner database. We examined new-onset menstrual bleeding disturbances using a Cox regression comparing vaccinated to unvaccinated women, where women contributed follow-up time as unvaccinated until the day of vaccination. In addition, we conducted a self-controlled case-series analysis, and a sensitivity analysis excluding all those who remained unvaccinated throughout the pandemic, to evaluate the role of unmeasured confounding. We observed an increased risk of several menstrual bleeding disturbances after vaccination against Covid-19, ranging from an adjusted HR (aHR) of 1.18 (95% CI: 1.04, 1.33) for intermenstrual bleeding to 1.29 (95% CI: 1.23, 1.36) for irregular/frequent menstrual periods. However, estimates were fully attenuated when excluding women who remained unvaccinated at the end of follow-up (aHRs between 0.97 and 1.08). No differences were identified according to vaccine dose or type. Our self-controlled case series analysis confirmed no increased risk after a first dose of vaccination, though there was a slightly increased risk of menstrual bleeding disturbances from 61 days after vaccination with dose 2. In conclusion, the modestly increased risk of menstrual bleeding disturbances after Covid-19 vaccination appeared to reflect a role of unmeasured confounding by women who never received Covid-19 vaccinations, as associations did not remain when risk after vaccination were compared to risk before vaccination among ever vaccinated women.