Causal inference from randomized trials in social epidemiology

Social epidemiology is the study of relations between social factors and health status in populations. Although recent decades have witnessed a rapid development of this research program in scope and sophistication, causal inference has proven to be a persistent dilemma due to the natural assignment of exposure level based on unmeasured attributes of individuals, which may lead to substantial confounding. Some optimism has been expressed about randomized social interventions as a solution to this long-standing inferential problem. We review the causal inference problem in social epidemiology, and the potential for causal inference in randomized social interventions. Using the example of a currently on-going intervention that randomly assigns families to non-poverty housing, we review the limitations to causal inference even under experimental conditions and explain which causal effects become identifiable. We note the benefit of using the randomized trial as a conceptual model, even for design and interpretation of observational studies in social epidemiology.     

观察性研究方法与医院资料的统计分析

目的通过个案分析,说明如何采用观察性研究的方法对医院资料进行统计分析.方法正确的研究设计与搜集客观准确的基础数据,明确观察对象所能推论的总体,设立对比组,通过分层分析及率的标准化等方法进行组间差别比较.结果医院资料不同于试验数据.由于医院资料通常不能随机分组,并存在混杂,不能简单地用假设检验方法推论因果关系.结论在对医院资料进行统计分析时,应遵循观察性研究的原则和方法,以避免可能存在的信息偏倚、选择偏倚和混杂偏倚.     

Do pollution time-series studies contain uncontrolled or residual confounding by risk factors for acute health events?

Acute health effects from air pollution are based largely on weak associations identified in time-series studies comparing daily air pollution levels to daily mortality. Much of this mortality is due to cardiovascular disease. Time-series studies have many potential limitations, but are not thought to be confounded by traditional cardiovascular risk factors (e.g., smoking status or hypertension) because these chronic risk factors are not obviously associated with daily pollution levels. However, acute psychobehavioral variants of these risk factors (e.g., smoking patterns and episodes of stress on any given day) are plausible confounders for the associations observed in time-series studies, given that time-series studies attempt to predict acute rather than chronic health outcomes. There is a fairly compelling literature on the strong link between cardiovascular events and daily "triggers" such as stress. Stress-related triggers are plausibly associated with daily pollution levels through surrogate stressors such as ambient temperature, daily workload, local traffic congestion, or other correlates of air pollution. For example, variables such as traffic congestion and industrial activity increase both stress-related health events and air pollution, suggesting the potential for classical confounding. Support for this argument is illustrated through examples of the well-demonstrated relationship between emotional stress and heart attack/stroke.     

协变量在因果中介效应分析中的作用

本文目的是介绍因果中介效应分析的理论基础以及结合一个实例采用SAS实现因果中介效应分析的具体方法。因果中介效应分析的理论基础包括基本概念以及定义因果中介效应的反事实框架。实例是关于父母提供的鼓励性环境是否会影响儿童的认知发展,分别采用传统的多重线性回归分析、不考虑协变量和考虑协变量的因果中介效应分析,通过比较3种分析方法所得到的结果,得出如下结论：①当资料中存在中介变量时,不适合采用传统的多重线性回归分析取代因果中介效应分析;②当资料中存在协变量时,不适合在忽视协变量的条件下进行因果中介效应分析。     

Log-linear model selections in a rural dental health study

This field study sought to measure the effects of dental delivery and school-based, dental health education on use of dental health care by children in grades K-6. We attempted to control for two potential confounding factors by an approximate randomization of children into treatment groups with stratification on grade and initial oral disease levels. A backward elimination log-linear model selection procedure for the 5-factor classification permitted tests for higher-order interaction, namely effect-modification, confounding and collapsibility. We found that the effect of dental health education on use of dental care depended on the mode of dental delivery.     

The prognostic role of perioperative allogeneic blood transfusions in gastric cancer patients undergoing curative resection: A systematic review and meta-analysis of non-randomized,adjusted studies

The impact of allogeneic perioperative blood transfusions (APTs) on the prognosis of gastric cancer patients undergoing curative-intent gastrectomy is still a highly debated topic. Two meta-analyses were published in 2015, and new studies report conflicting results.A literature review was conducted using PubMed, Scopus, the Cochrane Central Register of Controlled Trials and the Cochrane Database of Systematic Reviews, updated to March 1, 2016. Thirty-eight non-randomized studies reporting data on overall survival (OS), disease-free survival (DFS), disease-specific survival (DSS) and postoperative complications (PCs) were included. An inverse variance random-effects meta-analysis was conducted.APTs showed an association with worse OS, DFS, DSS and an increased number of PCs. The hazard ratio (HR) for OS was 1.49, with a 95% confidence interval (95% CI) of 1.32–1.69 (p < .00001; Q-test p = .001, I-squared = 56%). After outlier exclusion, the HR for OS was 1.34 (95% CI = 1.23–1.45, p < .00001; Q-test p = .64, I-squared = 0%). The HR for DFS was 1.48 (95% CI = 1.18–1.86, p = .0007; Q-test p = .31, I-squared = 16%), and the HR for DSS was 1.66 (95% CI = 1.5–2.19, p = .0004; Q-test p = .96, I-squared = 0%). The odds ratio for PCs was 3.33 (95% CI = 2.10–5.29, p < .00001; Q-test p = .14, I-squared = 42%).This meta-analysis showed a significant association between transfusions and OS, DFS, DSS and PCs. The quality of the evidence was low. Aggregation, selection and selective reporting bias were detected. The biases shifted the results towards significance. Further studies using accurate adjustment methods are needed. Until such additional studies are performed, caution in administering transfusions and optimization of cancer patient blood management are warranted.     

Lifestyle,socioeconomic characteristics,and medical history of elderly persons who receive seasonal influenza vaccination in a tax-supported healthcare system

BackgroundObservational studies on effectiveness of influenza vaccination in the elderly are thought to be biased by healthier lifestyles and higher socioeconomic status among vaccinated vs. unvaccinated persons. We examined this hypothesis in a uniform tax-supported health care system with free-of-charge influenza vaccination to the elderly.MethodsWe conducted a cross-sectional study among Danes aged 65–79 years participating in a survey. We compared elderly persons with and without a recent (within six months) influenza vaccination in terms of (i) lifestyle and socioeconomic characteristics obtained from the survey and (ii) health factors including medical history provided by Danish registries. We compared the prevalence of study variables among vaccinated and unvaccinated persons using age- and sex-adjusted prevalence ratios (aPRs) with 95% confidence intervals (CIs).ResultsAmong the 4237 elderly persons completing the survey, 1718 (41%) had received an influenza vaccination. Vaccinated persons had more comorbidity than unvaccinated persons (aPR for high comorbidity level: 1.51 95% CI 1.24–1.84), were less likely to never have smoked (aPR: 0.88, 95% CI 0.80–0.97), and had a higher prevalence of physical inactivity (aPR: 1.08, 95% CI 1.03–1.13). Levels of education and income were similar in the two groups. Vaccinated persons had a higher prevalence of major physical limitations (aPR: 1.40, 95% CI 1.17–1.66) and need for assistance with activities of daily living (aPR: 1.29, 95% CI 1.13–1.47).ConclusionElderly influenza vaccinated persons were not healthier in terms of lifestyle and burden of disease, did not have a higher socioeconomic status, and were more frail than unvaccinated persons.     

Assessment of Confounding and Interaction using the Mantel-Haenszel Risk Estimation Method

The association between an exposure of interest (risk factor) and a disease may be confounded by the action of other separate factors as well as by interactions between risk factors exerting an impct. Crude measures of effect may be misleading in such situations. Levels of the potential confounding factor could be estimated using stratified analysis. Uniformity of the stratum-specific effect estimates can be assessed by performing chi-square tests for heterogeneity. If the effect is uniform across strata, we can calculate a pooled adjusted summary estimate of the effect using the Mantel-Haenzel (M-H) method. Confidence intervals for the adjusted estimate and the M-H chi-square test are calculated to assess the significance. If the effect is not uniform (presence of interaction), we report stratum-specific estimates, confidence intervals and chi-square for each estimate. In the present paper, assessment of the level of confounding and interaction between risk factors is illustrated using a case-control study of lung cancer conducted at the Regional Cancer Centre, Trivandrum.     

Characterising socio-economic inequalities in exposure to air pollution: a comparison of socio-economic markers and scales of measurement

This study examines traffic-related air pollution in London in relation to area- and individual-level socio-economic position (SEP). Mean air pollution concentrations were generally higher in postcodes of low SEP as classified by small-area markers of deprivation (Index of Multiple Deprivation (IMD) domains) and by the postcode-level ACORN geodemographic marker. There were exceptions, however, including reversed directions of associations in central London and for SEP markers relating to education. ACORN predicted air pollution independently of IMD and explained additional variation at the postcode level, indicating the potential value of using both markers in air pollution epidemiology studies. By contrast, after including IMD and ACORN there remained little relationship between air pollution and individual-level SEP or smoking, suggesting limited residual socio-economic confounding in epidemiological studies with comprehensive area-level adjustment.