Search

Feminization of male and juvenile fish because of exposure to estrogens or estrogenic chemicals in effluents from central wastewater treatment plants (WWTPs) is a worldwide issue of concern. Intersex and induction of the female yolk protein, vitellogenin, in male and juvenile fish are robust biomarkers for estrogenic exposure, and feminized fish have been observed downstream of WWTP outlets in many countries. Danish central WWTPs reduce effluent estrogenicity effectively by advanced sewage treatment, and feminizations have not been observed downstream of central WWTP outlets. However, between 2000 and 2004, investigations of Danish streams not receiving sewage from central WWTPs revealed a high variation in vitellogenin concentrations of male juvenile brown trout (Salmo trutta); some individuals had high concentrations, probably as a result of estrogenic point sources, and the plasma concentration was >50 ng mL −1 in 79% of the juvenile males. The streams were reinvestigated in 2010 to 2016, and the average male level had decreased to a hitherto unseen baseline level; in 2010 only 0.7% (one individual) of the males had a vitellogenin concentration >50 ng mL −1, which could indicate that the estrogenicity of the streams decreased after 2004. We examined possible estrogenic sources in streams unaffected by central WWTP effluents, and found that the reduced vitellogenin levels are most likely explained by a national effort to improve on-site wastewater treatment in scattered houses not connected to central WWTPs. Environ Toxicol Chem 2018;37:839–845.

Many exposures of epidemiological interest are time varying, and the values of potential confounders may change over time leading to time varying confounding. The aim of many longitudinal studies is to estimate the causal effect of a time varying exposure on an outcome that requires adjusting for time varying confounding. Time varying confounding affected by previous exposure often occurs in practice, but it is usually adjusted for by using conventional analytical methods such as time dependent Cox regression, random effects models, or generalised estimating equations, which are known to provide biased effect estimates in this setting. This article explains time varying confounding affected by previous exposure and outlines three causal methods proposed to appropriately adjust for this potential bias: inverse-probability-of-treatment weighting, the parametric G formula, and G estimation.

The high-dimensional propensity score is a semiautomated variable selection algorithm that can supplement expert knowledge to improve confounding control in nonexperimental medical studies utilizing electronic healthcare databases. Although the algorithm can be used to generate hundreds of patient-level variables and rank them by their potential confounding impact, it remains unclear how to select the optimal number of variables for adjustment. We used plasmode simulations based on empirical data to discuss and evaluate data-adaptive approaches for variable selection and prediction modeling that can be combined with the high-dimensional propensity score to improve confounding control in large healthcare databases. We considered approaches that combine the high-dimensional propensity score with Super Learner prediction modeling, a scalable version of collaborative targeted maximum-likelihood estimation, and penalized regression. We evaluated performance using bias and mean squared error (MSE) in effect estimates. Results showed that the high-dimensional propensity score can be sensitive to the number of variables included for adjustment and that severe overfitting of the propensity score model can negatively impact the properties of effect estimates. Combining the high-dimensional propensity score with Super Learner was the most consistent strategy, in terms of reducing bias and MSE in the effect estimates, and may be promising for semiautomated data-adaptive propensity score estimation in high-dimensional covariate datasets.

Confounding is an important source of bias, but it is often misunderstood. We consider how confounding occurs and how to address confounding using examples. Study results are confounded when the effect of the exposure on the outcome, mixes with the effects of other risk and protective factors for the outcome. This problem arises when these factors are present to different degrees among the exposed and unexposed study participants, but not all differences between the groups result in confounding. Thinking about an ideal study where all of the population of interest is exposed in one universe and is unexposed in a parallel universe helps to distinguish confounders from other differences. In an actual study, an observed unexposed population is chosen to stand in for the unobserved parallel universe. Differences between this substitute population and the parallel universe result in confounding. Confounding by identified factors can be addressed analytically and through study design, but only randomization has the potential to address confounding by unmeasured factors. Nevertheless, a given randomized study may still be confounded. Confounded study results can lead to incorrect conclusions about the effect of the exposure of interest on the outcome.

© 2017, The International Biometric Society Complex interplay between genetic and environmental factors characterizes the etiology of many diseases. Modeling gene–environment (GxE) interactions is often challenged by the unknown functional form of the environment term in the true data-generating mechanism. We study the impact of misspecification of the environmental exposure effect on inference for the GxE interaction term in linear and logistic regression models. We first examine the asymptotic bias of the GxE interaction regression coefficient, allowing for confounders as well as arbitrary misspecification of the exposure and confounder effects. For linear regression, we show that under gene–environment independence and some confounder-dependent conditions, when the environment effect is misspecified, the regression coefficient of the GxE interaction can be unbiased. However, inference on the GxE interaction is still often incorrect. In logistic regression, we show that the regression coefficient is generally biased if the genetic factor is associated with the outcome directly or indirectly. Further, we show that the standard robust sandwich variance estimator for the GxE interaction does not perform well in practical GxE studies, and we provide an alternative testing procedure that has better finite sample properties.

Background Veno-arterial extracorporeal membrane oxygenation (VA ECMO) is an effective rescue therapy for severe cardiorespiratory failure, but morbidity and mortality are high. We hypothesised that survival decreases with longer VA ECMO treatment. We examined the Extracorporeal Life Support Organization (ELSO) registry for a relationship between VA ECMO duration and in-hospital mortality, and covariates including indication for support. Methods All VA runs from the ELSO database from 2002 to 2012 were extracted. Multiple runs and non-VA runs were excluded. Runs were categorized into diagnostic groups. Logistic regression for analysis of the effect of duration on outcome, and multivariate regression for diagnosis and other baseline factors were performed. Non-linear models including piecewise logistic models were fitted. Results There were 2699 runs analysed over 14,747 days. Logistic regression analysis of the effect of duration on outcome, and multivariate regression analysis of diagnosis and other baseline factors were performed. In-hospital survival was 41.4% (95% CI 39.6–43.3). 75% of patients were supported for less than 1 week and 96% for less than 3 weeks. Median duration (4 days IQR 2.0–6.8) was greater in survivors (4.1 (IQR 2.5–6.7) vs 3.8 (IQR 1.7–7.0) p = 0.002). The final multivariate model demonstrated increasing survival to day 4 (OR 1.53 (95% CI 1.37–1.71) p

Background In mediation analysis if unmeasured confounding is present, the estimates for the direct and mediated effects may be over or under estimated. Most methods for the sensitivity analysis of unmeasured confounding in mediation have focused on the mediator-outcome relationship. Results The Umediation R package enables the user to simulate unmeasured confounding of the exposure-mediator, exposure-outcome, and mediator-outcome relationships in order to see how the results of the mediation analysis would change in the presence of unmeasured confounding. We apply the Umediation package to the Genetic Epidemiology of Chronic Obstructive Pulmonary Disease (COPDGene) study to examine the role of unmeasured confounding due to population stratification on the effect of a single nucleotide polymorphism (SNP) in the CHRNA5/3/B4 locus on pulmonary function decline as mediated by cigarette smoking. Conclusions Umediation is a flexible R package that examines the role of unmeasured confounding in mediation analysis allowing for normally distributed or Bernoulli distributed exposures, outcomes, mediators, measured confounders, and unmeasured confounders. Umediation also accommodates multiple measured confounders, multiple unmeasured confounders, and allows for a mediator-exposure interaction on the outcome. Umediation is available as an R package at https://github.com/SharonLutz/Umediation A tutorial on how to install and use the Umediation package is available in the Additional file 1. Electronic supplementary material The online version of this article (doi:10.1186/s12859-017-1749-y) contains supplementary material, which is available to authorized users.

Sensitivity analysis is useful in assessing how robust an association is to potential unmeasured or uncontrolled confounding. This article introduces a new measure called the "E-value," which is related to the evidence for causality in observational studies that are potentially subject to confounding. The E-value is defined as the minimum strength of association, on the risk ratio scale, that an unmeasured confounder would need to have with both the treatment and the outcome to fully explain away a specific treatment-outcome association, conditional on the measured covariates. A large E-value implies that considerable unmeasured confounding would be needed to explain away an effect estimate. A small E-value implies little unmeasured confounding would be needed to explain away an effect estimate. The authors propose that in all observational studies intended to produce evidence for causality, the E-value be reported or some other sensitivity analysis be used. They suggest calculating the E-value for both the observed association estimate (after adjustments for measured confounders) and the limit of the confidence interval closest to the null. If this were to become standard practice, the ability of the scientific community to assess evidence from observational studies would improve considerably, and ultimately, science would be strengthened.

It is generally agreed that motor performance and executive functioning (EF) are intertwined. As the literature on this issue concerning preschool children is scarce, we examined the relationship between motor performance and parent-rated EF in a sample of 3- to 5-year-old children with different levels of motor skill proficiency, while controlling for age, gender, socio-economic status (SES), and attention-deficit-hyperactivity disorder (ADHD) symptomatology. EF was reported by parents of 153 children (mean age 4years 1months, SD 8months; 75 male) by means of the Behaviour Rating Inventory of Executive Function-Preschool version (BRIEF-P). Parent-reported ADHD symptoms were assessed using the Hyperactivity-Inattention subscale of the Strengths and Difficulties Questionnaire3-4. In addition, the children performed the Movement Assessment Battery for Children-2 (MABC-2). Several weak to moderate relationships were found between the MABC-2 Total Score and the EF subscales. Once other variables such as age, gender, SES, and ADHD symptomatology were taken into account, the only BRIEF-P subscale that was associated with the MABC-2 Total Score was the Working Memory subscale. Compared to their typically developing peers, children who are at risk for motor coordination difficulties (⩽the 16th percentile on the MABC-2) performed poorly on the Working Memory subscale, which confirms the results of the regression analyses. The at risk group also performed significantly worse on the Planning/Organize subscale, however. This is one of the first studies investigating the relationship between motor performance and parent-rated EF in such a young age group. It shows that the relationship between motor performance and EF in young children is complex and may be influenced by the presence of confounding variables such as ADHD symptomatology.

Background: Objectives: To develop a healthy user index to serve as a method of confounding adjustment in future observational studies of preventive therapies. Methods: A large administrative database of patients with type 2 diabetes was split in half randomly, yielding derivation and validation cohorts. Influenza vaccination was used as a ‘prototypical marker’ of a healthy user. In our derivation cohort, we fitted a mixed effects logistic regression model, and a points-based system was used to construct the index. The healthy user index was then evaluated in the validation cohort. Results: Overall, 13% had received the influenza vaccination. In the derivation cohort (n= 914 732), the healthy user index ranged from 0 to 91 with a mean of 41.6 (SD 12.9). When applied to the validation cohort (n= 913 231), the index ranged from 0 to 96 (mean 41.6, SD 12.9) and significantly predicted influenza vaccination with a c-statistic of 0.605, suggesting moderate discrimination ability. Conclusion: Our healthy user index combined age, sex and healthy behaviors to predict healthy users within administrative datasets. This index score may allow for better adjustment of healthy user bias in health services research; however, external validation is further required. Key Words: Administrative Data Uses, Bias, Biostatistical Methods, Observational Data

Numerous models have been developed to quantify the combined effect of various risk factors to predict either risk of developing breast cancer, risk of carrying a high-risk germline genetic mutation, specifically in the BRCA1 and BRCA2 genes, or the risk of both. These breast cancer risk models can be separated into those that utilize mainly hormonal and environmental factors and those that focus more on hereditary risk. Given the wide range of models from which to choose, understanding what each model predicts, the populations for which each is best suited to provide risk estimations, the current validation and comparative studies that have been performed for each model, and how to apply them practically is important for clinicians and researchers seeking to utilize risk models in their practice. This review provides a comprehensive guide for those seeking to understand and apply breast cancer risk models by summarizing the majority of existing breast cancer risk prediction models including the risk factors they incorporate, the basic methodology in their development, the information each provides, their strengths and limitations, relevant validation studies, and how to access each for clinical or investigative purposes.

The study of mediation has a long tradition in the social sciences and a relatively more recent one in epidemiology. The first school is linked to path analysis and structural equation models (SEMs), while the second is related mostly to methods developed within the potential outcomes approach to causal inference. By giving model-free definitions of direct and indirect effects and clear assumptions for their identification, the latter school has formalized notions intuitively developed in the former and has greatly increased the flexibility of the models involved. However, through its predominant focus on nonparametric identification, the causal inference approach to effect decomposition via natural effects is limited to settings that exclude intermediate confounders. Such confounders are naturally dealt with (albeit with the caveats of informality and modeling inflexibility) in the SEM framework. Therefore, it seems pertinent to revisit SEMs with intermediate confounders, armed with the formal definitions and (parametric) identification assumptions from causal inference. Here we investigate: 1) how identification assumptions affect the specification of SEMs, 2) whether the more restrictive SEM assumptions can be relaxed, and 3) whether existing sensitivity analyses can be extended to this setting. Data from the Avon Longitudinal Study of Parents and Children (1990-2005) are used for illustration.

Calls for the adoption of complex systems approaches, including agent-based modeling, in the field of epidemiology have largely centered on the potential for such methods to examine complex disease etiologies, which are characterized by feedback behavior, interference, threshold dynamics, and multiple interacting causal effects. However, considerable theoretical and practical issues impede the capacity of agent-based methods to examine and evaluate causal effects and thus illuminate new areas for intervention. We build on this work by describing how agent-based models can be used to simulate counterfactual outcomes in the presence of complexity. We show that these models are of particular utility when the hypothesized causal mechanisms exhibit a high degree of interdependence between multiple causal effects and when interference (i.e., one person's exposure affects the outcome of others) is present and of intrinsic scientific interest. Although not without challenges, agent-based modeling (and complex systems methods broadly) represent a promising novel approach to identify and evaluate complex causal effects, and they are thus well suited to complement other modern epidemiologic methods of etiologic inquiry.

Studies of weight and mortality sometimes state that the mortality relative risks for obesity from nonsmokers are valid estimates of the relative risks for obesity in both smokers and nonsmokers. Extending this idea, several influential articles have used relative risks for obesity from nonsmokers and attributable fraction methods for unadjusted risks to estimate attributable fractions of deaths in the entire population (smokers and nonsmokers combined). However, stratification by smoking is a form of adjustment for confounding. Simplified examples show that the use of relative risks from only 1 stratum to estimate attributable fractions, without incorporating data on the stratification variable, gives incorrect results for the entire population. Even if the mortality relative risks for obesity from nonsmokers are indeed valid in both smokers and nonsmokers, these relative risks nonetheless need to be treated as adjusted relative risks for the purpose of calculating attributable fractions for the whole sample.

Imagine an oncologist knocks on your door with the following problem: she wants to know how much she could reduce mortality among her bone marrow transplant patients by prescribing a new drug that prevents graft-versus-host disease, a side effect of allogeneic marrow transplantation.1 While graft-versus-host disease is associated in observational studies with an increased risk of mortality, it also reduces the risk of leukemia relapse – thus, any drug that prevents graft-versus-host disease may have the very undesirable side effect of increasing the rate of relapse.2 She wants to compare the mortality in her cohort with what mortality would be in that same cohort if they had taken this new drug. We cannot answer this question with a regression model because leukemia relapse is a risk factor for mortality and subsequent graft-versus-host disease and it will also decrease the incidence of subsequent relapse (i.e. relapse is a confounder affected by exposure).3, 4 However, we can answer this question using the g-formula. The g-formula is an analytic tool for estimating standardized outcome distributions using covariate (exposure and confounders) specific estimates of the outcome distribution.5The g-formula can be used to estimate familiar measures of association, such as the hazard ratio. In the current paper, we address the oncologist’s question: we compare observed mortality in our cohort with the expected mortality in that cohort under the new treatment. Epidemiologists often use regression models (for example, the Cox proportional hazards model) to adjust for confounding; this is equivalent to estimating stratum-specific hazard ratios and then averaging the information-weighted hazard ratios. When some of those confounders are also causal intermediates, this amounts to adjusting away some of the effect of exposure.6, 7 The g-formula works differently: first, one finds weighted averages of the stratum-specific hazards, and then those averaged (standardized) hazards are combined in a summary hazard ratio. Thus, bias resulting from time-varying covariates that can be both confounders and causal intermediates is a shortcoming of using regression models to control for confounding, rather than a general principle of observational data analysis.8, 9 The g-formula is a tool that overcomes this shortcoming, but its use in the literature has been sparse – we could find only 9 examples using observational data.8, 10-17 We hypothesize that the dearth of software packages and lack of useful, yet simple, examples of the g-formula have been the main barriers to broader use. We show how the g-formula can be used with standard software tools that many epidemiologists already employ, and we illustrate it using publicly-available data from a small cohort study with accompanying SAS code in an eAppendix. We illustrate how we can estimate the net (total) effect of a hypothetical treatment to prevent graft-versus-host disease on mortality and compare the g-formula approach with a regression approach. The g-formula (as with any statistical method) relies on making assumptions in order to make sense of the complex processes underlying the data. We discuss possible ways to assess how well we meet the assumptions as well as the robustness of the g-formula to violations of these assumptions.

BACKGROUND Recent data show a falling cancer mortality in the general population without a similar shift in immigrant outcomes, leading to a greater cancer burden and mortality for immigrants. Our aims were to compare perceived patterns of care in immigrants and native-born cancer patients. PATIENTS AND METHODS This was a hospital-based sample of first-generation immigrants and Australian-born Anglo patients in the first year following diagnosis. It was restricted to Chinese, Arabic, or Greek speakers. Eligible participants, recruited via 16 oncology clinics, were over 18, with cancer (any type or stage), and having commenced treatment at least 1 month previously. Five hundred and seventy-one CALD patients (comprising 145 Arabic, 248 Chinese, and 178 Greek) and a control group of 274 Anglo-Australian patients participated. RESULTS Immigrants had difficulty communicating with the doctor (73% versus 29%) and understanding the health system (38% versus 10%). Differences were found in 'difficulty knowing who to see' (P = 0.0002), 'length of time to confirm diagnosis' (P = 0.04), wanting more choice about a specialist and hospital (P < 0.0001); being offered the opportunity to see a counselor (P < 0.0001); and actually seeing one (P < 0.0001). There were no significant self-reported differences regarding how cancer was detected, time to see a health professional, or type first seen; however, immigrants reported difficulty knowing who to see. Previous studies showed differences in patterns of care according to socioeconomic status (SES) and educational level. Despite adjusting for age, sex, education, marital status, SES, time since diagnosis, and type of cancer, we did not find significant differences. Instead, we found that understanding of the health system and confidence understanding English were important factors. CONCLUSIONS This study confirmed that immigrants with cancer perceive an inferior quality of cancer care. We highlight potentially modifiable factors including assistance in navigating the health system, translated information, and cultural competency training for health professionals.

Propensity scores are widely used to control for confounding when estimating the effect of a binary treatment in observational studies. They have been generalized to ordinal and continuous treatments in the recent literature. Following the definition of propensity function and its parameterizations (called the propensity parameter in this paper) proposed by Imai and van Dyk, we explore sufficient conditions for selecting propensity parameters to control for confounding for continuous treatments in the context of regression-based adjustment in linear models. Typically, investigators make parametric assumptions about the form of the dose-response function for a continuous treatment. Such assumptions often allow the analyst to use only a subset of the propensity parameters to control confounding. When the treatment is the only predictor in the structural, that is, causal model, it is sufficient to adjust only for the propensity parameters that characterize the expectation of the treatment variable or its functional form. When the structural model includes selected baseline covariates other than the treatment variable, those baseline covariates, in addition to the propensity parameters, must also be adjusted in the model. We demonstrate these points with an example estimating the dose-response relationship for the effect of erythropoietin on hematocrit level in patients with end-stage renal disease.

Using a reduced subset of SNPs in a linear mixed model can improve power for genome-wide association studies, yet this can result in insufficient correction for population stratification. We propose a hybrid approach using principal components that does not inflate statistics in the presence of population stratification and improves power over standard linear mixed models.

We consider several possible interpretations of the “effect of race” when regressions are run with race as an exposure variable, controlling also for various confounding and mediating variables. When adjustment is made for socioeconomic status early in a person’s life, we discuss under what contexts the regression coefficients for race can be interpreted as corresponding to the extent to which a racial inequality would remain if various socioeconomic distributions early in life across racial groups could be equalized. When adjustment is also made for adult socioeconomic status, we note how the overall racial inequality can be decomposed into the portion that would be eliminated by equalizing adult socioeconomic status across racial groups and the portion of the inequality that would remain even if adult socioeconomic status across racial groups were equalized. We also discuss a stronger interpretation of the “effect of race” (stronger in terms of assumptions) involving the joint effects of race-associated physical phenotype (e.g. skin color), parental physical phenotype, genetic background and cultural context when such variables are thought to be hypothetically manipulable and if adequate control for confounding were possible. We discuss some of the challenges with such an interpretation. Further discussion is given as to how the use of selected populations in examining racial disparities can additionally complicate the interpretation of the effects.

Purpose Geriatric frailty is a common syndrome of older adults that is characterized by increased vulnerability to adverse health outcomes and influences treatment choice. Pharmacoepidemiologic studies in older adults that rely on administrative claims data are limited by confounding due to unmeasured frailty. A claim-based frailty score may be useful to minimize confounding by frailty in such databases. We provide an overview of definitions and measurement of frailty, evaluated the claim-based models of frailty in literature, and recommend ways to improve frailty adjustment in claims analysis.

In cohort and case-control studies, confounding that arises as a result of differences in the distribution of determinants of the outcome between exposure groups leading to non-exchangeability are addressed by restriction, matching or with statistical models. In case-only studies, this issue is addressed by comparing each individual with his/herself. Although case-only designs use self-matching and only include individuals who develop the outcome of interest, issues of non-exchangeability are identical to those that arise in traditional case-control and cohort studies. In this review, we describe one type of case-only design, the case-crossover design, and discuss how the concept of exchangeability can be used to understand issues of confounding, carryover effects, period effects and selection bias in case-crossover studies.

Natural direct and indirect effects formalize traditional notions of mediation analysis into a rigorous causal framework and have recently received considerable attention in epidemiology and in social sciences. Sufficient conditions for the identification of natural direct effects were formulated by Judea Pearl under a nonparametric structural equations model, which assumes certain independencies between potential outcomes. A common situation in epidemiology is that a confounder of the mediator-outcome relationship is itself affected by the exposure, in which case natural direct effects fail to be nonparametrically identified without additional assumptions, even under Pearl's nonparametric structural equations model. In this article, we show that when a single binary confounder of the mediator is affected by the exposure, the natural direct effect is nonparametrically identified under the model, assuming monotonicity about the effect of the exposure on the confounder. A similar result is shown to hold for a vector of binary confounders of the mediator under a certain independence assumption about the confounders. Finally, we show that natural direct effects are more generally identified if there is no additive mean interaction between the mediator and the confounders of the mediator affected by exposure. When correct, this latter assumption is particularly appealing because it does not require monotonicity of effects of the exposure. In addition, it places no restriction on the nature of the confounders of the mediator, which can be continuous or polytomous.

Association studies form the backbone of biomedical research, with almost every effort in the field ultimately boiling down to a comparison between groups, coupled with some form of statistical test intended to determine whether or not any observed difference is more or less than would be expected by chance. Unfortunately, although the paradigm is powerful and frequently effective, it is often forgotten that false positive association can easily arise if there is any bias or systematic difference in the way in which study subjects are selected into the considered groups. To protect against such confounding, researchers generally try to match cases and controls for extraneous variables thought to correlate with the exposures of interest. However, if seemingly homogenously distributed exposures are actually more heterogeneous than appreciated, then matching may be inadequate and false positive results can still arise. In this review, we will illustrate these fundamental issues by considering the previously proposed relationship between month of birth and multiple sclerosis. This much discussed but false positive association serves as a reminder of just how heterogeneous even easily measured environmental risk factors can be, and how easily case control studies can be confounded by seemingly minor differences in ascertainment. Electronic supplementary material The online version of this article (doi:10.1007/s00415-014-7241-y) contains supplementary material, which is available to authorized users.

Data from randomized controlled trials (RCTs) provide the strongest evidence for establishing relations between exposures, including dietary exposures, and health outcomes. However, not all diet and health outcome relations can be practically or ethically evaluated by using RCTs; therefore, many dietary recommendations are supported by evidence primarily from observational data, particularly those from prospective cohort studies. Although such evidence is of critical importance, limitations are often underappreciated by nutrition scientists and policymakers. This editorial review is intended to 1) highlight some of these limitations of observational evidence for diet-disease relations, including imprecise exposure quantification, collinearity among dietary exposures, displacement/substitution effects, healthy/unhealthy consumer bias, residual confounding, and effect modification; and 2) advocate for greater caution in the communication of dietary recommendations for which RCT evidence of clinical event reduction after dietary intervention is not available.

Unobserved confounding can seldom be ruled out with certainty in nonexperimental studies. Negative controls are sometimes used in epidemiologic practice to detect the presence of unobserved confounding. An outcome is said to be a valid negative control variable to the extent that it is influenced by unobserved confounders of the exposure effects on the outcome in view, although not directly influenced by the exposure. Thus, a negative control outcome found to be empirically associated with the exposure after adjustment for observed confounders indicates that unobserved confounding may be present. In this paper, we go beyond the use of control outcomes to detect possible unobserved confounding and propose to use control outcomes in a simple but formal counterfactual-based approach to correct causal effect estimates for bias due to unobserved confounding. The proposed control outcome calibration approach is developed in the context of a continuous or binary outcome, and the control outcome and the exposure can be discrete or continuous. A sensitivity analysis technique is also developed, which can be used to assess the degree to which a violation of the main identifying assumption of the control outcome calibration approach might impact inference about the effect of the exposure on the outcome in view.

The purpose of this study was to examine the influence of explanatory and confounding variables on health-related quality of life after accounting for response shift, measurement bias and response shift in measurement using structural equation modeling. Hypertensive patients with coronary artery disease randomized to anti-hypertensive treatment, completed the ShortForm-36 questionnaire at both baseline and 1 year (n = 788). Three measurement biases were found and all three were considered as response shift in measurement. Older patients reported worse scores for both physical functioning (PF) and role-physical at baseline and 1 year later compared to younger patients; and males reported better PF than females after conditioning on the latent trait of general physical health. Before controlling for response shift, patients' PF scores were not statistically different over time; however, PF scores significantly improved (p0.01) after controlling for recalibration response shift. Assessment of how patients perceive their change in health-related quality of life over time is warranted.

Mammographic density measurements are associated with risk of breast cancer. Few studies have investigated the concurrent associations of mammographic dense and nondense areas, body mass index (weight (kg)/height (m)(2)), and ages at mammogram and diagnosis with breast cancer risk. We conducted a matched, case-control study nested within the Melbourne Collaborative Cohort Study (cohort recruitment in 1990-1994 and follow-up until 2007) to estimate the associations between these factors and breast cancer risk under alternative causal models. Mammographic dense area was positively associated with risk, and the strength of this association was only slightly influenced by the choice of the causal model (relative risk per 1 standard deviation = 1.50, 95% confidence interval: 1.32, 1.70). Mammographic nondense area was inversely associated with risk under the assumption that fat in the body and fat in the breast cause breast cancer through independent mechanisms (relative risk per 1 standard deviation = 0.75, 95% confidence interval: 0.65, 0.86), whereas it was not associated with risk under the assumption that they are both proxies of adiposity. Knowledge about the biological mechanisms regulating the role played by mammographic nondense area and body fat on breast cancer risk is essential to better estimate their impacts on individual risk.

Researchers have identified environmental risks that predict subsequent psychological and medical problems. Based on these correlational findings, researchers have developed and tested complex developmental models and have examined biological moderating factors (e.g., gene–environment interactions). In this context, we stress the critical need for researchers to use family-based, quasi-experimental designs when trying to integrate genetic and social science research involving environmental variables because these designs rigorously examine causal inferences by testing competing hypotheses. We argue that sibling comparison, offspring of twins or siblings, in vitro fertilization designs, and other genetically informed approaches play a unique role in bridging gaps between basic biological and social science research. We use studies on maternal smoking during pregnancy to exemplify these principles.

This paper addresses strategies for selecting variables for adjustment in non-experimental comparative effectiveness research and uses causal graphs to illustrate the causal network that relates treatment to outcome. Variables in the causal network take on multiple structural forms. Adjustment for a common cause pathway between treatment and outcome can remove confounding, whereas adjustment for other structural types may increase bias. For this reason, variable selection would ideally be based on an understanding of the causal network; however, the true causal network is rarely known. Therefore, we describe more practical variable selection approaches based on background knowledge when the causal structure is only partially known. These approaches include adjustment for all observed pretreatment variables thought to have some connection to the outcome, all known risk factors for the outcome, and all direct causes of the treatment or the outcome. Empirical approaches, such as forward and backward selection and automatic high-dimensional proxy adjustment, are also discussed. As there is a continuum between knowing and not knowing the causal, structural relations of variables, we recommend addressing variable selection in a practical way that involves a combination of background knowledge and empirical selection and that uses high-dimensional approaches. This empirical approach can be used to select from a set of a priori variables based on the researcher's knowledge to be included in the final analysis or to identify additional variables for consideration. This more limited use of empirically derived variables may reduce confounding while simultaneously reducing the risk of including variables that may increase bias.

Preinfluenza periods have been used to test for uncontrolled confounding in studies of influenza vaccine effectiveness, but some authors have claimed that confounding differs in preinfluenza and influenza periods. We tested this claim by comparing estimates of the vaccine-mortality association during the 2009/2010 influenza year, when there was essentially no circulation of seasonal influenza in the United States, and 2007/2008, a typical influenza year. We pooled data on seniors (adults aged ≥65 years) from 7 US managed care organizations that participated in the Vaccine Safety Datalink Project. We defined influenza vaccination, all-cause mortality, and potential confounders from administrative databases. We quantified the vaccine-mortality association using Cox regression. During 2007/2008, the adjusted hazard ratio was 0.44 prior to influenza season, 0.62 during influenza season, and 0.71 after influenza season. A similar pattern was observed during 2009/2010, when any effect of seasonal influenza vaccine observed during all time periods must have resulted from confounding: 0.65 during the autumn, 0.80 during the winter, and 0.84 during the summer. In a year with minimal seasonal influenza, we found no evidence that confounding in autumn preinfluenza periods is qualitatively different from confounding in winter. This supports the use of preinfluenza periods as control time periods in studies of influenza vaccine effectiveness.

Background: Maternal Smoking During Pregnancy (SDP) has consistently been associated with increased risk of attention-deficit/hyperactivity disorder (ADHD) in offspring, but recent studies indicate that this association might be due to unmeasured familial confounding. Methods: A total of 813,030 individuals born in Sweden between 1992 and 2000 were included in this nationwide population-based cohort study. Data on maternal SDP and ADHD diagnosis were obtained from national registers and patients were followed up from the age of 3 to the end of 2009. Hazard Ratios (HRs) were estimated using stratified Cox regression models. Cousin and sibling data were used to control for unmeasured familial confounding. Results: At the population level maternal SDP predicted ADHD in offspring (HRModerateSDP = 1.89; HRHighSDP = 2.50). This estimate gradually attenuated toward the null when adjusting for measured confounders (HRModerateSDP = 1.62; HRHighSDP = 2.04), unmeasured confounders shared within the extended family (i.e., cousin comparison) (HRModerateSDP = 1.45; HRHighSDP = 1.69), and unmeasured confounders within the nuclear family (i.e., sibling comparison) (HRModerateSDP = 0.88; HRHighSDP = 0.84). Conclusions: Our results suggest that the association between maternal SDP and offspring ADHD are due to unmeasured familial confounding. Keywords: Maternal smoking during pregnancy, attention-deficit/hyperactivity disorder, confounding, sibling comparisons.

Objectives This review assessed the confounding effect of one traffic-related exposure (noise or air pollutants) on the association between the other exposure and cardiovascular outcomes. Methods A systematic review was conducted with the databases Medline and Embase. The confounding effects in studies were assessed by using change in the estimate with a 10 % cutoff point. The influence on the change in the estimate of the quality of the studies, the exposure assessment methods and the correlation between road noise and air pollutions were also assessed. Results Nine publications were identified. For most studies, the specified confounders produced changes in estimates

Instrumental variable analysis is a technique commonly used in the social sciences to provide evidence that a treatment causes an outcome, as contrasted with evidence that a treatment is merely associated with differences in an outcome. To extract such strong evidence from observational data, instrumental variable analysis exploits situations where some degree of randomness affects how patients are selected for a treatment. An instrumental variable is a characteristic of the world that leads some people to be more likely to get the specific treatment we want to study but does not otherwise change those patients' outcomes. This seminar explains, in nonmathematical language, the logic behind instrumental variable analyses, including several examples. It also provides three key questions that readers of instrumental variable analyses should ask to evaluate the quality of the evidence. (1) Does the instrumental variable lead to meaningful differences in the treatment being tested? (2) Other than through the specific treatment being tested, is there any other way the instrumental variable could influence the outcome? (3) Does anything cause patients to both receive the instrumental variable and receive the outcome?

BACKGROUND Prolonged, intensive treatment regimens often disrupt families of children with cancer. Siblings are at increased risk for distress, but factors underlying this risk have received limited empirical attention. In this study, the authors examined associations between the family context and sibling distress. METHODS Siblings of children with cancer (ages 8-18 years; N = 209) and parents (186 mothers and 70 fathers) completed measures of sibling distress, family functioning, parenting, and parent post-traumatic stress. Associations between sibling distress and each family risk factor were evaluated. Then, family risks were considered simultaneously by calculating cumulative family risk index scores. RESULTS After controlling for sociodemographic covariates, greater sibling distress was associated with more sibling-reported problems with family functioning and parental psychological control, lower sibling-reported maternal acceptance, and lower paternal self-reported acceptance. When risk factors were considered together, the results supported a quadratic model in which associations between family risk and sibling distress were stronger at higher levels of risk. CONCLUSIONS The current findings support a contextual model of sibling adjustment to childhood cancer in which elevated distress is predicted by family risk factors, both alone and in combination. Cancer 2013;119:2503-2510. © 2013 American Cancer Society.

Background : Recent research suggests that the Bayesian paradigm may be useful for modeling biases in epidemiological studies, such as those due to misclassification and missing data. We used Bayesian methods to perform sensitivity analyses for assessing the robustness of study findings to the potential effect of these two important sources of bias. Methods : We used data from a study of the joint associations of radiotherapy and smoking with primary lung cancer among breast cancer survivors. We used Bayesian methods to provide an operational way to combine both validation data and expert opinion to account for misclassification of the two risk factors and missing data. For comparative purposes we considered a "full model" that allowed for both misclassification and missing data, along with alternative models that considered only misclassification or missing data, and the naive model that ignored both sources of bias. Results : We identified noticeable differences between the four models with respect to the posterior distributions of the odds ratios that described the joint associations of radiotherapy and smoking with primary lung cancer. Despite those differences we found that the general conclusions regarding the pattern of associations were the same regardless of the model used. Overall our results indicate a nonsignificantly decreased lung cancer risk due to radiotherapy among nonsmokers, and a mildly increased risk among smokers. Conclusions : We described easy to implement Bayesian methods to perform sensitivity analyses for assessing the robustness of study findings to misclassification and missing data.

The validity of epidemiologic findings depends on the completeness and accuracy with which covariates are measured and taken into account. For some variables such as comorbid conditions or concurrent therapies, it is often not possible to distinguish between missing data and absence of the condition. Operationally, missing information for such variables is usually assumed to indicate that the condition is not present.1 In cohort studies conducted within healthcare utilization databases, the sensitivity with which such variables are detected depends in large part on the duration of time over which each subject had been observed before follow-up began.2 Because subjects will vary in terms of the duration of available baseline information, the investigator is faced with a choice: 1) utilize covariate information based on all available baseline data for each subject, 2) utilize covariate information based on a fixed baseline window of time that is shared by all subjects (often chosen as 6 months or 1 year). We term these options the “all-available” and “fixed window” approaches. It can be assumed that under the all-available paradigm, confounders will be identified with equal or greater sensitivity than under a fixed window approach, which should serve to reduce residual confounding.3-4 On the other hand, the duration of available baseline data may be associated with subject covariates, subsequent exposure and outcome (each either directly or mediated through other unmeasured confounders), so that the all-available approach could introduce bias through differential confounder ascertainment. Consider the following motivating example. Investigators are interested in estimating the comparative effectiveness of new use of anti-platelet agent A versus a standard anti-platelet agent B on 30-day stroke risk following a transient ischemic attack among Medicaid beneficiaries in secondary claims data. Diabetes is an important potential confounder. The source cohort comprises patients who have a claim involving a diagnosis code for transient ischemic attack, and who subsequently filled a prescription for anti-platelet medication A or B (index fill). Potential subjects are excluded if they demonstrate claims for anti-platelet agents before the index fill, or if they do not have 6 months of available data before the index fill over which to document no prior exposure. Some patients meeting entry criteria may have the minimum requisite 6 months of baseline data, whereas others may have many years of baseline data. The duration of available data likely relates to observable characteristics (eg, socioeconomic status, disability) as well as unobservable ones (eg, frailty) that themselves are associated with diabetic status, choice of anti-platelet agent and risk of subsequent stroke. To define diabetes, the investigator must determine whether to identify claims suggesting diabetes using all available data for each patient, or to restrict to data from the collectively shared 6-month baseline period. The latter approach results in uniform observation periods for study participants, but that comes at the cost of ignoring some data available for some patients. We sought to explore the tradeoffs inherent in this choice through a simulation study that compared bias reduction under these two approaches, and to determine the conditions under which one approach may be preferable to the other. Specifically, we considered the following circumstance: 1) the effect of interest was that of a dichotomous exposure on a dichotomous outcome; 2) the observed exposure—outcome association was confounded by a single observed dichotomous covariate; 3) all study subjects included in the analysis had available data for at least the 6-month period preceding exposure, and variable amounts of lead-up time before that period. We then compared the two approaches in the presence of an additional unmeasured covariate that was associated with exposure, outcome, duration of historical follow up and frequency of medical contact. Of note, we have considered confounding based on a chronic disease (or analogous) conception; results do not pertain to covariates that confound on a temporary or transient basis (eg, acute conditions).

Uses of the propensity score to obtain estimates of causal effect have been investigated thoroughly under assumptions of linearity and additivity of exposure effect. When the outcome variable is binary relationships such as collapsibility, valid for the linear model, do not always hold. This article examines uses of the propensity score when both exposure and outcome are binary variables and the parameter of interest is the marginal odds ratio. We review stratification and matching by the propensity score when calculating the Mantel–Haenszel estimator and show that it is consistent for neither the marginal nor conditional odds ratio. We also investigate a marginal odds ratio estimator based on doubly robust estimators and summarize its performance relative to other recently proposed estimators under various conditions, including low exposure prevalence and model misspecification. Finally, we apply all estimators to a case study estimating the effect of Medicare plan type on the quality of care received by African-American breast cancer patients.

Measurement error threatens public health by producing bias in estimates of the population impact of environmental exposures. Quantitative methods to account for measurement bias can improve public health decision making. We summarize traditional and emerging methods to improve inference under a standard perspective, in which the investigator estimates an exposure-response function, and a policy perspective, in which the investigator directly estimates population impact of a proposed intervention. Under a policy perspective, the analyst must be sensitive to errors in measurement of factors that modify the effect of exposure on outcome, must consider whether policies operate on the true or measured exposures, and may increasingly need to account for potentially dependent measurement error of two or more exposures affected by the same policy or intervention. Incorporating approaches to account for measurement error into such a policy perspective will increase the impact of environmental epidemiology.

BACKGROUND: Breast cancer comprises several molecular subtypes with different prognoses and possibly different etiology. Reproductive and hormonal factors are associated with breast cancer overall, and with luminal subtypes, but the associations with other subtypes are unclear. We used data from a national screening program to conduct a large nested case-control study. METHODS: We conducted a nested case-control study on participants in the Norwegian Breast Cancer Screening Program in 2006 - 2014. There was information on estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor receptor 2 (HER2) for 4748 cases of breast cancer. Breast cancer subtypes were defined as luminal A-like (ER+ PR+ HER2-), luminal B-like (ER+ PR- HER2- or ER+ PR+/PR-HER2+), HER2-positive (ER- PR- HER2+) and triple-negative (ER- PR- HER2-). Conditional logistic regression was used to estimate odds ratios (ORs) of breast cancer associated with age at first birth, number of pregnancies, oral contraceptive use, intrauterine devices and menopausal hormone therapy. Analyses were adjusted for age, body mass index, education, age at menarche, number of pregnancies and menopausal status. RESULTS: Number of pregnancies was inversely associated with relative risk of luminal-like breast cancers (p-trend ≤0.02), and although not statistically significant, with HER2-positive (OR = 0.60, 95% CI 0.31-1.19) and triple-negative cancer (OR = 0.70, 95% CI 0.41-1.21). Women who had ≥4 pregnancies were at >40% lower risk of luminal-like and HER2-positive cancers than women who had never been pregnant. However, there was a larger discrepancy between tumor subtypes with menopausal hormone use. Women who used estrogen and progesterone therapy (EPT) had almost threefold increased risk of luminal A-like cancer (OR = 2.92, 95% CI 2.36-3.62) compared to never-users, but were not at elevated risk of HER2-positive (OR = 0.88, 95% CI 0.33-2.30) or triple-negative (OR = 0.92, 95% CI 0.43 - 1.98) subtypes. CONCLUSIONS: Reproductive factors were to some extent associated with all subtypes; the strongest trends were with luminal-like subtypes. Hormone therapy use was strongly associated with risk of luminal-like breast cancer, and less so with risk of HER2-positive or triple-negative cancer. There are clearly some, but possibly limited, etiologic differences between subtypes, with the greatest contrast between luminal A-like and triple-negative subtypes. TRIAL REGISTRATION: Not applicable.

The amount of family financial resources available in early life influences child health and development. Using data from the 2000 and 2007 waves of the Indonesian Family Life Survey, we estimated the associations of early-life poverty (at age

Objective To examine the association between a history of periodontitis (PD) and incident systemic lupus erythematosus (SLE) Methods We used 2003–2012 claims data from the Taiwanese National Health Insurance Database to identify 7,204 incident SLE patients during 2007–2012 as the study group, along with randomly selecting 72,040 non-SLE patients matched (1:10) for age, gender, and first diagnosis date (index date) as the control group. The correlation between PD and SLE risk was estimated using conditional logistic regression analysis, after making adjustments for confounders (including a history of diabetes and number of non-PD related dental visits before the index date). To evaluate the effects of PD severity and the lag time which occurred since the last PD visit on SLE development, odds ratios (ORs) with 95% confidence intervals (CIs) were calculated for subgroups of patients with PD according to their number of visits, cumulative cost and also the time gaps between their last PD-related visit and the index date. Results A statistically significant association between a history of PD and newly diagnosed SLE was observed (OR, 1.21; 95% CI, 1.14–1.28; p-value

Background: Ischaemic heart disease (IHD) incidence is elevated in people reporting psychological distress. The extent to which this relationship is causal or related to reverse causality-that is, undiagnosed disease causing distress-is unclear. We quantified the relationship between psychological distress and IHD, with consideration of confounding and undiagnosed disease. Methods: Questionnaire data (2006–2009) from 151 811 cardiovascular disease-free and cancer-free Australian general population members aged ≥45years (45 and Up Study) were linked to hospitalisation and mortality data, to December 2013. A two-stage approach estimated HRs for incident IHD (IHD-related hospitalisation or death) for low (Kessler-10 scores: 10–

Observational studies almost always have bias because prognostic factors are unequally distributed between patients exposed or not exposed to an intervention. The standard approach to dealing with this problem is adjusted or stratified analysis. Its principle is to use measurement of risk factors to create prognostically homogeneous groups and to combine effect estimates across groups.The purpose of this Users' Guide is to introduce readers to fundamental concepts underlying adjustment as a way of dealing with prognostic imbalance and to the basic principles and relative trustworthiness of various adjustment strategies.One alternative to the standard approach is propensity analysis, in which groups are matched according to the likelihood of membership in exposed or unexposed groups. Propensity methods can deal with multiple prognostic factors, even if there are relatively few patients having outcome events. However, propensity methods do not address other limitations of traditional adjustment: investigators may not have measured all relevant prognostic factors (or not accurately), and unknown factors may bias the results.A second approach, instrumental variable analysis, relies on identifying a variable associated with the likelihood of receiving the intervention but not associated with any prognostic factor or with the outcome (other than through the intervention); this could mimic randomization. However, as with assumptions of other adjustment approaches, it is never certain if an instrumental variable analysis eliminates bias.Although all these approaches can reduce the risk of bias in observational studies, none replace the balance of both known and unknown prognostic factors offered by randomization.

Association studies among admixed populations pose many challenges including confounding of genetic effects due to population substructure and heterogeneity due to different patterns of linkage disequilibrium (LD). We use simulations to investigate controlling for confounding by indicators of global ancestry and the impact of including a covariate for local ancestry. In addition, we investigate the use of an interaction term between a single-nucleotide polymorphism (SNP) and local ancestry to capture heterogeneity in SNP effects. Although adjustment for global ancestry can control for confounding, additional adjustment for local ancestry may increase power when the induced admixture LD is in the opposite direction as the LD in the ancestral population. However, if the induced LD is in the same direction, there is the potential for reduced power because of overadjustment. Furthermore, the inclusion of a SNP by local ancestry interaction term can increase power when there is substantial differential LD between ancestry populations. We examine these approaches in genome-wide data using the University of Southern California's Children's Health Study investigating asthma risk. The analysis highlights rs10519951 (P = 8.5 × 10(-7)), a SNP lacking any evidence of association from a conventional analysis (P = 0.5).