Search

BACKGROUND: Cardiovascular disease (CVD) is the leading cause of death worldwide. It has been known for some considerable time that radiation is associated with excess risk of CVD. A recent systematic review of radiation and CVD highlighted substantial inter-study heterogeneity in effect, possibly a result of confounding or modifications of radiation effect by non-radiation factors, in particular by the major lifestyle/environmental/medical risk factors and latent period. METHODS: We assessed effects of confounding by lifestyle/environmental/medical risk factors on radiation-associated CVD and investigated evidence for modifying effects of these variables on CVD radiation dose-response, using data assembled for a recent systematic review. RESULTS: There are 43 epidemiologic studies which are informative on effects of adjustment for confounding or risk modifying factors on radiation-associated CVD. Of these 22 were studies of groups exposed to substantial doses of medical radiation for therapy or diagnosis. The remaining 21 studies were of groups exposed at much lower levels of dose and/or dose rate. Only four studies suggest substantial effects of adjustment for lifestyle/environmental/medical risk factors on radiation risk of CVD; however, there were also substantial uncertainties in the estimates in all of these studies. There are fewer suggestions of effects that modify the radiation dose response; only two studies, both at lower levels of dose, report the most serious level of modifying effect. CONCLUSIONS: There are still large uncertainties about confounding factors or lifestyle/environmental/medical variables that may influence radiation-associated CVD, although indications are that there are not many studies in which there are substantial confounding effects of these risk factors.

Systematically missing data in distributed data networks presents practical and methodological challenges. Failure to handle it appropriately can bias statistical inference. Multiple imputations can be used to address systematic missingness. However, when data from different study sites cannot be pooled into a unified file, conventional imputation approaches become unavailable due to the absence of a basis for imputation. To address such challenges, we introduce an imputation method based on conditional quantiles – conditional quantile imputation (CQI) – which involves four steps: (i) estimating 99 quantiles for the systematically missing variable in studies with observed data; (ii) deriving a weighted average of regression coefficients across studies and transmitting it to sites with systematically missing data; (iii) imputing the systematically missing values based on observed data and the set of regression coefficients from step ii; and (iv) combining estimates of the substantive outcome model across imputations using Rubin's rules. We evaluate CQI in different simulation scenarios and illustrate it with an applied data example. We conclude that CQI can be a suitable approach for the imputation of systematically missing data when data from multiple studies cannot be pooled. [ABSTRACT FROM AUTHOR]

Background: Methods for matching in longitudinal cohort studies, such as sequential stratification and time-varying propensity scores, facilitate causal inferences in the context of time-dependent treatments that are not randomized where patient eligibility or treatment status changes over time. The tradeoffs in available approaches have not been compared previously, so we compare two methods using simulations based on a retrospective cohort of patients eligible for weight loss surgery, some of whom received it. Methods: This study compares matching completeness, bias, coverage, and precision among three approaches to longitudinal matching: (1) time-varying propensity scores (tvPS), (2) sequential stratification that matches exactly on all covariates used in tvPS (SS-Full) and (3) sequential stratification that exact matches on a subset of covariates (SS-Selected). These comparisons are made in the context of a deep sampling frame (50:1) and a shallow sampling frame (5:1) of eligible comparators. A simulation study was employed to estimate the relative performance of these approaches. Results: In 1,000 simulations each, tvPS retained more than 99.9% of treated patients in both the deep and shallow sampling frames, while a smaller proportion of treated patients were retained for SS-Full (91.6%) and SS-Selected (98.2%) in the deep sampling frame. In the shallow sampling frame, sequential stratification retained many fewer treated patients (73.9% SS-Full, 92.0% SS-Selected) than tvPS yet coverage, precision and bias were comparable for tvPS, SS-Full and SS-Selected in the deep and shallow sampling frames. Conclusion: Time-varying propensity scores have comparable performance to sequential stratification in terms of coverage, bias, and precision, with superior match completeness. While performance was generally comparable across methods, greater match completeness makes tvPS an attractive option for longitudinal matching studies where external validity is highly valued. [ABSTRACT FROM AUTHOR]

Background: Assessing maternal health care utilisation is imperative for the health of both mother and her child. Maternal education is an important determinant in subsequent maternal health care usage, according to research. There is a dearth of research on the causal relationship between maternal education and maternal health services as well as examining the performance of different propensity score methods for estimating absolute effects. Therefore, this study aims to estimate the effect of maternal education on usage of maternal health services minimising the confounding effect and to compare different propensity score approaches to estimate absolute effects of maternal education on usage of maternal health services. Methods: We used data from a cross-sectional study conducted by icddr,b in Bangladesh. A total of 1300 recently delivered mothers were included in this study. We used different propensity score (PS) methods to estimate the causal effect of maternal education on receiving maternal health services, including PS matching, PS weighting, covariate adjustment using PS, as well as used modified log-Poisson regression with and without multiple covariates. Results: The study revealed highly imbalanced possible confounding factors for mother's educational level, which might lead to erroneous estimates. All methods indicated a significant effect of higher maternal education, whether measured as a continuous scale or a binary variable (secondary or higher vs. below secondary) on maternal health services, significantly increased the likelihood of receiving four or more ANC, delivered at facility, PNC within 42 days and receiving all maternal health services compared to the mothers with lower level of education. The PS weighting provided precise estimates with a low range of confidence interval. Conclusions: The results provided important insights on how well these techniques worked to reduce effect of confounding variables and achieve precise estimates. Propensity score weighting method performed better in terms of providing more precise estimates with a narrower range of confidence intervals, indicating that this method may be a reliable approach for estimating the causal effect of maternal education on maternal health service utilisation. However, careful consideration should be given to selecting the most appropriate method. [ABSTRACT FROM AUTHOR]

From early in the coronavirus disease 2019 (COVID‐19) pandemic, there was interest in using machine learning methods to predict COVID‐19 infection status based on vocal audio signals, for example, cough recordings. However, early studies had limitations in terms of data collection and of how the performances of the proposed predictive models were assessed. This article describes how these limitations have been overcome in a study carried out by the Turing‐RSS Health Data Laboratory and the UK Health Security Agency. As part of the study, the UK Health Security Agency collected a dataset of acoustic recordings, SARS‐CoV‐2 infection status and extensive study participant meta‐data. This allowed us to rigorously assess state‐of‐the‐art machine learning techniques to predict SARS‐CoV‐2 infection status based on vocal audio signals. The lessons learned from this project should inform future studies on statistical evaluation methods to assess the performance of machine learning techniques for public health tasks. [ABSTRACT FROM AUTHOR]

Least absolute shrinkage and selection operator (LASSO) regression is widely used for large-scale propensity score (PS) estimation in health-care database studies. In these settings, previous work has shown that undersmoothing (overfitting) LASSO PS models can improve confounding control, but it can also cause problems of nonoverlap in covariate distributions. It remains unclear how to select the degree of undersmoothing when fitting large-scale LASSO PS models to improve confounding control while avoiding issues that can result from reduced covariate overlap. Here, we used simulations to evaluate the performance of using collaborative-controlled targeted learning to data-adaptively select the degree of undersmoothing when fitting large-scale PS models within both singly and doubly robust frameworks to reduce bias in causal estimators. Simulations showed that collaborative learning can data-adaptively select the degree of undersmoothing to reduce bias in estimated treatment effects. Results further showed that when fitting undersmoothed LASSO PS models, the use of cross-fitting was important for avoiding nonoverlap in covariate distributions and reducing bias in causal estimates. [ABSTRACT FROM AUTHOR]

Analysis of variance and linear models is undoubtedly one of the most useful statistical contributions to experimental and observational science. With the ability to characterize a system through multivariate responses, these methods have emerged to be general tools regardless of response dimensionality. Contemporary methods for establishing statistical inference, such as ANOVA simultaneous component analysis (ASCA), are based on Monte Carlo sampling; however, a flat uniform resampling scheme may violate the structure of the uncertainty for unbalanced designs as well as for observational data. In this work, we provide permutation strategies for inferential testing for unbalanced designs including interaction models and establish nonuniform randomization based on the concept of propensity score matching. Lastly, we provide a general method for modelling continuous covariates based on kernel smoothers. All methods are characterized on their ability to provide unbiased Type I error results. [ABSTRACT FROM AUTHOR]

Aim: Non‐randomized studies on bariatric surgery have reported large reductions in mortality within 6‐12 months after surgery compared with non‐surgical patients. It is unclear whether these findings are the result of bias. Study Design and Setting: We searched PubMed to identify all non‐randomized studies investigating the effect of bariatric surgery on all‐cause mortality compared with non‐surgical patients. We assessed these studies for potential confounding and time‐related biases. We conducted bias analyses to quantify the effect of these biases. Results: We identified 21 cohort studies that met our inclusion criteria. Among those, 11 were affected by immortal time bias resulting from the misclassification or exclusion of relevant follow‐up time. Five studies were subject to potential confounding bias because of a lack of adjustment for body mass index (BMI). All studies used an inadequate comparator group that lacked indications for bariatric surgery. Bias analyses to correct for potential confounding from BMI shifted the effect estimates towards the null [reported hazard ratio (HR): 0.78 vs. bias‐adjusted HR: 0.92]. Bias analyses to correct for the presence of immortal time also shifted the effect estimates towards the null (adjustment for 2‐year wait time: reported HR: 0.57 vs. bias‐adjusted HR: 0.81). Conclusion: Several important sources of bias were identified in non‐randomized studies of the effectiveness of bariatric surgery versus non‐surgical comparators on mortality. Future studies should ensure that confounding by BMI is accounted for, considering the choice of the comparator group, and that the design or analysis avoids immortal time bias from the misclassification or exclusion. [ABSTRACT FROM AUTHOR]

Abstract Background Assessing maternal health care utilisation is imperative for the health of both mother and her child. Maternal education is an important determinant in subsequent maternal health care usage, according to research. There is a dearth of research on the causal relationship between maternal education and maternal health services as well as examining the performance of different propensity score methods for estimating absolute effects. Therefore, this study aims to estimate the effect of maternal education on usage of maternal health services minimising the confounding effect and to compare different propensity score approaches to estimate absolute effects of maternal education on usage of maternal health services. Methods We used data from a cross-sectional study conducted by icddr,b in Bangladesh. A total of 1300 recently delivered mothers were included in this study. We used different propensity score (PS) methods to estimate the causal effect of maternal education on receiving maternal health services, including PS matching, PS weighting, covariate adjustment using PS, as well as used modified log-Poisson regression with and without multiple covariates. Results The study revealed highly imbalanced possible confounding factors for mother's educational level, which might lead to erroneous estimates. All methods indicated a significant effect of higher maternal education, whether measured as a continuous scale or a binary variable (secondary or higher vs. below secondary) on maternal health services, significantly increased the likelihood of receiving four or more ANC, delivered at facility, PNC within 42 days and receiving all maternal health services compared to the mothers with lower level of education. The PS weighting provided precise estimates with a low range of confidence interval. Conclusions The results provided important insights on how well these techniques worked to reduce effect of confounding variables and achieve precise estimates. Propensity score weighting method performed better in terms of providing more precise estimates with a narrower range of confidence intervals, indicating that this method may be a reliable approach for estimating the causal effect of maternal education on maternal health service utilisation. However, careful consideration should be given to selecting the most appropriate method.

Abstract Background Methods for matching in longitudinal cohort studies, such as sequential stratification and time-varying propensity scores, facilitate causal inferences in the context of time-dependent treatments that are not randomized where patient eligibility or treatment status changes over time. The tradeoffs in available approaches have not been compared previously, so we compare two methods using simulations based on a retrospective cohort of patients eligible for weight loss surgery, some of whom received it. Methods This study compares matching completeness, bias, coverage, and precision among three approaches to longitudinal matching: (1) time-varying propensity scores (tvPS), (2) sequential stratification that matches exactly on all covariates used in tvPS (SS-Full) and (3) sequential stratification that exact matches on a subset of covariates (SS-Selected). These comparisons are made in the context of a deep sampling frame (50:1) and a shallow sampling frame (5:1) of eligible comparators. A simulation study was employed to estimate the relative performance of these approaches. Results In 1,000 simulations each, tvPS retained more than 99.9% of treated patients in both the deep and shallow sampling frames, while a smaller proportion of treated patients were retained for SS-Full (91.6%) and SS-Selected (98.2%) in the deep sampling frame. In the shallow sampling frame, sequential stratification retained many fewer treated patients (73.9% SS-Full, 92.0% SS-Selected) than tvPS yet coverage, precision and bias were comparable for tvPS, SS-Full and SS-Selected in the deep and shallow sampling frames. Conclusion Time-varying propensity scores have comparable performance to sequential stratification in terms of coverage, bias, and precision, with superior match completeness. While performance was generally comparable across methods, greater match completeness makes tvPS an attractive option for longitudinal matching studies where external validity is highly valued.

Abstract Background In microbiome disease association studies, it is a fundamental task to test which microbes differ in their abundance between groups. Yet, consensus on suitable or optimal statistical methods for differential abundance testing is lacking, and it remains unexplored how these cope with confounding. Previous differential abundance benchmarks relying on simulated datasets did not quantitatively evaluate the similarity to real data, which undermines their recommendations. Results Our simulation framework implants calibrated signals into real taxonomic profiles, including signals mimicking confounders. Using several whole meta-genome and 16S rRNA gene amplicon datasets, we validate that our simulated data resembles real data from disease association studies much more than in previous benchmarks. With extensively parametrized simulations, we benchmark the performance of nineteen differential abundance methods and further evaluate the best ones on confounded simulations. Only classic statistical methods (linear models, the Wilcoxon test, t-test), limma, and fastANCOM properly control false discoveries at relatively high sensitivity. When additionally considering confounders, these issues are exacerbated, but we find that adjusted differential abundance testing can effectively mitigate them. In a large cardiometabolic disease dataset, we showcase that failure to account for covariates such as medication causes spurious association in real-world applications. Conclusions Tight error control is critical for microbiome association studies. The unsatisfactory performance of many differential abundance methods and the persistent danger of unchecked confounding suggest these contribute to a lack of reproducibility among such studies. We have open-sourced our simulation and benchmarking software to foster a much-needed consolidation of statistical methodology for microbiome research.

Abstract Background Cluster randomized trials (CRTs) are randomized trials where randomization takes place at an administrative level (e.g., hospitals, clinics, or schools) rather than at the individual level. When the number of available clusters is small, researchers may not be able to rely on simple randomization to achieve balance on cluster-level covariates across treatment conditions. If these cluster-level covariates are predictive of the outcome, covariate imbalance may distort treatment effects, threaten internal validity, lead to a loss of power, and increase the variability of treatment effects. Covariate-constrained randomization (CR) is a randomization strategy designed to reduce the risk of imbalance in cluster-level covariates when performing a CRT. Existing methods for CR have been developed and evaluated for two- and multi-arm CRTs but not for factorial CRTs. Methods Motivated by the BEGIN study—a CRT for weight loss among patients with pre-diabetes—we develop methods for performing CR in 2 × 2 factorial cluster randomized trials with a continuous outcome and continuous cluster-level covariates. We apply our methods to the BEGIN study and use simulation to assess the performance of CR versus simple randomization for estimating treatment effects by varying the number of clusters, the degree to which clusters are associated with the outcome, the distribution of cluster level covariates, the size of the constrained randomization space, and analysis strategies. Results Compared to simple randomization of clusters, CR in the factorial setting is effective at achieving balance across cluster-level covariates between treatment conditions and provides more precise inferences. When cluster-level covariates are included in the analyses model, CR also results in greater power to detect treatment effects, but power is low compared to unadjusted analyses when the number of clusters is small. Conclusions CR should be used instead of simple randomization when performing factorial CRTs to avoid highly imbalanced designs and to obtain more precise inferences. Except when there are a small number of clusters, cluster-level covariates should be included in the analysis model to increase power and maintain coverage and type 1 error rates at their nominal levels.

Unmeasured confounding is a key challenge for causal inference. In this paper, we establish a framework for unmeasured confounding adjustment with negative control variables. A negative control outcome is associated with the confounder but not causally affected by the exposure in view, and a negative control exposure is correlated with the primary exposure or the confounder but does not causally affect the outcome of interest. We introduce an outcome confounding bridge function that depicts the relationship between the confounding effects on the primary outcome and the negative control outcome, and we incorporate a negative control exposure to identify the bridge function and the average causal effect. We also consider the extension to the positive control setting by allowing for the nonzero causal effect of the primary exposure on the control outcome. We illustrate our approach with simulations and apply it to a study about the short-term effect of air pollution on mortality. Although a standard analysis shows a significant acute effect of PM2.5 on mortality, our analysis indicates that this effect may be confounded, and after double negative control adjustment, the effect is attenuated toward zero.

Electronic health records and other sources of observational data are increasingly used for drawing causal inferences. The estimation of a causal effect using these data not meant for research purposes is subject to confounding and irregularly-spaced covariate-driven observation times affecting the inference. A doubly-weighted estimator accounting for these features has previously been proposed that relies on the correct specification of two nuisance models used for the weights. In this work, we propose a novel consistent multiply robust estimator and demonstrate analytically and in comprehensive simulation studies that it is more flexible and more efficient than the only alternative estimator proposed for the same setting. It is further applied to data from the Add Health study in the United States to estimate the causal effect of therapy counseling on alcohol consumption in American adolescents. [ABSTRACT FROM AUTHOR]

This article studies the inference of the regression coefficient matrix under multivariate response linear regressions in the presence of hidden variables. A novel procedure for constructing confidence intervals of entries of the coefficient matrix is proposed. Our method first uses the multivariate nature of the responses by estimating and adjusting the hidden effect to construct an initial estimator of the coefficient matrix. By further deploying a low-dimensional projection procedure to reduce the bias introduced by the regularization in the previous step, a refined estimator is proposed and shown to be asymptotically normal. The asymptotic variance of the resulting estimator is derived with closed-form expression and can be consistently estimated. In addition, we propose a testing procedure for the existence of hidden effects and provide its theoretical justification. Both our procedures and their analyses are valid even when the feature dimension and the number of responses exceed the sample size. Our results are further backed up via extensive simulations and a real data analysis. for this article are available online. [ABSTRACT FROM AUTHOR]

Background High-dimensional propensity scoring (HDPS) is a method for empirically identifying potential confounders within large healthcare databases such as administrative claims data. However, this method has not yet been applied to large national health surveys such as the National Health and Aging Trends Study (NHATS), an ongoing nationally representative survey of older adults in the United States and important resource in gerontology research. Methods In this Research Practice article, we present an overview of HDPS and describe the specific data transformation steps and analytic considerations needed to apply it to national health surveys. We applied HDPS within NHATS to investigate the association between self-reported visual difficulty and incident dementia, comparing HDPS to conventional confounder selection methods. Results Among 7 207 dementia-free NHATS Wave 1 respondents, 528 (7.3%) had self-reported visual difficulty. In an unadjusted discrete time proportional hazards model accounting for the complex survey design of NHATS, self-reported visual difficulty was strongly associated with incident dementia (odds ratio [OR] 2.34, 95% confidence interval [CI]: 1.95–2.81). After adjustment for standard investigator-selected covariates via inverse probability weighting, the magnitude of this association decreased, but evidence of an association remained (OR 1.44, 95% CI: 1.11–1.85). Adding 75 HDPS-prioritized variables to the investigator-selected propensity score model resulted in further attenuation of the association between visual impairment and dementia (OR 0.94, 95% CI: 0.70–1.23). Conclusions HDPS can be successfully applied to national health surveys such as NHATS and may improve confounder adjustment. We hope developing this framework will encourage future consideration of HDPS in this setting. [ABSTRACT FROM AUTHOR]

Simpson’s paradox (SP) is a statistical phenomenon where the association between two variables reverses, disappears, or emerges, after conditioning on a third variable. It has been proposed (by, e.g., Judea Pearl) that SP should be analyzed using the framework of graphical causal models (i.e., causal DAGs) in which SP is diagnosed as a symptom of confounding bias. This paper contends that this confounding-based analysis cannot fully capture SP: there are cases of SP that cannot be explained away in terms of confounding. Previous works have argued that some cases of SP do not require causal analysis at all. Despite being a logically valid counterexample, we argue that this type of cases poses only a limited challenge to Pearl’s analysis of SP. In our view, a more powerful challenge to Pearl comes from cases of SP that do require causal analysis but can arise without confounding. We demonstrate with examples that accidental associations due to genetic drift, the use of inappropriate aggregate variables as causes, and interactions between units (i.e., inter-unit causation) can all give rise to SP of this type. The discussion is also extended to the amalgamation paradox (of which SP is a special form) which can occur due to the use of non-collapsible association measures, in the absence of confounding. [ABSTRACT FROM AUTHOR]

Observational studies are frequently used in clinical research to estimate the effects of treatments or exposures on outcomes. To reduce the effects of confounding when estimating treatment effects, covariate balancing methods are frequently implemented. This study evaluated, using extensive Monte Carlo simulation, several methods of covariate balancing, and two methods for propensity score estimation, for estimating the average treatment effect on the treated using a hazard ratio from a Cox proportional hazards model. With respect to minimizing bias and maximizing accuracy (as measured by the mean square error) of the treatment effect, the average treatment effect on the treated weighting, fine stratification, and optimal full matching with a conventional logistic regression model for the propensity score performed best across all simulated conditions. Other methods performed well in specific circumstances, such as pair matching when sample sizes were large (n = 5000) and the proportion treated was < 0.25. Statistical power was generally higher for weighting methods than matching methods, and Type I error rates were at or below the nominal level for balancing methods with unbiased treatment effect estimates. There was also a decreasing effective sample size with an increasing number of strata, therefore for stratification-based weighting methods, it may be important to consider fewer strata. Generally, we recommend methods that performed well in our simulations, although the identification of methods that performed well is necessarily limited by the specific features of our simulation. The methods are illustrated using a real-world example comparing beta blockers and angiotensin-converting enzyme inhibitors among hypertensive patients at risk for incident stroke. [ABSTRACT FROM AUTHOR]

Background Perceived discrimination in health care settings can have adverse consequences on mental health in minority groups. However, the association between perceived discrimination and mental health is prone to unmeasured confounding. The study aims to quantitatively evaluate the influence of unmeasured confounding in this association, using g-estimation. Methods In a predominantly African American cohort, we applied g-estimation to estimate the association between perceived discrimination and mental health, adjusted and unadjusted for measured confounders. Mental health was measured using clinical diagnoses of anxiety, depression and bipolar disorder. Perceived discrimination was measured as the number of patient-reported discrimination events in health care settings. Measured confounders included demographic, socioeconomic, residential and health characteristics. The influence of confounding was denoted as α 1 from g-estimation. We compared α 1 for measured and unmeasured confounding. Results Strong associations between perceived discrimination in health care settings and mental health outcomes were observed. For anxiety, the odds ratio (95% confidence interval) unadjusted and adjusted for measured confounders were 1.30 (1.21, 1.39) and 1.26 (1.17, 1.36), respectively. The α 1 for measured confounding was –0.066. Unmeasured confounding with α 1 = 0.200 ⁠ , which was over three times that of measured confounding, corresponds to an odds ratio of 1.12 (1.01, 1.24). Similar results were observed for other mental health outcomes. Conclusion Compared with measured confounding, unmeasured that was three times measured confounding was not enough to explain away the association between perceived discrimination and mental health, suggesting that this association is robust to unmeasured confounding. This study provides a novel framework to quantitatively evaluate unmeasured confounding. [ABSTRACT FROM AUTHOR]

Associations of anticoagulation with primary endpoints in longitudinal studies are impacted by selection bias and time-varying covariates (e.g. comorbidities). We demonstrate how time-varying covariates and selection bias influence association estimates between anticoagulation and health-related quality of life (HRQoL) in patients with atrial fibrillation. We performed a secondary analysis of the Atrial Fibrillation Follow-up Investigation of Rhythm Management trial quality of life substudy. Dichotomized warfarin use was ascertained at the study baseline, 2 months later, and annually for up to 6 years. HRQoL was measured at every time point using a self-reported ordinal 5-point Likert-scale (lower score and lower odds ratio represents better health-related quality of life). Static and time-varying covariates were ascertained throughout the study period. Confounder-adjusted generalized mixed model and generalized estimating equation regressions were used to demonstrate traditional association estimates between anticoagulation and HRQoL. Inverse probability of treatment and censorship weights were used to ascertain the influence of time-varying confounding and selection bias. Age-stratified analysis (age ≥ 70 years) evaluated for effect modification. 656 individuals were included in the analysis, 601 on warfarin at baseline. The association of warfarin use with better HRQoL over time strengthened when accounting for time-varying confounding and selection bias (OR 0.30, 95% CI 0.14–0.55) compared to traditional analyses (OR 0.61, 95% CI 0.38–0.97), and was most pronounced in those ≥ 70 years upon stratified analysis. Anticoagulation is associated with higher HRQoL in patients with atrial fibrillation, with time-varying confounding and selection bias likely influencing longitudinal estimates in anticoagulation-HRQoL research. [ABSTRACT FROM AUTHOR]

Propensity score methods are popular to control for confounding in observational biomedical studies of risk factors or medical treatments. This paper focused on aspects of propensity score methods that often remain undiscussed, including unmeasured confounding, missing data, variable selection, statistical efficiency, estimands, the positivity assumption, and predictive performance of the propensity score model. [ABSTRACT FROM AUTHOR]

Local statistical analysis has long been of interest to social and environmental scientists who analyze geographic data. Research into local spatial statistics experienced a step-change in the mid-1990s, which provided a large class of local statistical methods and models. The local Moran statistic is one commonly used local indicator of spatial association, able to detect both areas of similarity and observations that are very dissimilar from their surroundings. From this, many further local statistics have been developed to characterize spatial clusters and outliers. These statistics have seen limited adoption because they do not sufficiently model the relationships involved in confounded spatial data, where the analyst seeks to understand the local spatial structure of a given outcome variable that is influenced by one or more additional factors. Recent innovations used to do joint multivariate local analysis also do not model this kind of conditional local structure in data. This article provides tools to rigorously characterize confounded local inference and a new and different class of multivariate conditional local Moran statistics that can account for confounding. To do this, we return to the Moran scatterplot as the critical tool for local Moran-style covariance statistics. Extending this concept, a new method is available directly from a "Moran-form" multiple regression. We show the empirical and theoretical properties of this statistic, show how some existing heuristic approaches arise naturally from this framework, and show how the use of conditional inference can change interpretations in an empirical analysis of rent and housing stock in a rapidly changing neighborhood. [ABSTRACT FROM AUTHOR]

Evidence on the comparative effectiveness of osteoporosis treatments is heterogeneous. This may be attributed to different populations and clinical practice, but also to differing methodologies ensuring comparability of treatment groups before treatment effect estimation and the amount of residual confounding by indication. This study assessed the comparability of denosumab vs oral bisphosphonate (OBP) groups using propensity score (PS) methods and negative control outcome (NCO) analysis. A total of 280 288 women aged ≥50 yr initiating denosumab or OBP in 2011-2018 were included from the UK Clinical Practice Research Datalink (CPRD) and the Danish National Registries (DNR). Balance of observed covariates was assessed using absolute standardized mean difference (ASMD) before and after PS weighting, matching, and stratification, with ASMD >0.1 indicating imbalance. Residual confounding was assessed using NCOs with ≥100 events. Hazard ratio (HR) and 95%CI between treatment and NCO were estimated using Cox models. Presence of residual confounding was evaluated with 2 approaches (1) >5% of NCOs with 95% CI excluding 1, (2) >5% of NCOs with an upper CI <0.75 or lower CI >1.3. The number of imbalanced covariates before adjustment (CPRD 22/87; DNR 18/83) decreased, with 2%–11% imbalance remaining after weighting, matching, or stratification. Using approach 1, residual confounding was present for all PS methods in both databases (≥8% of NCOs), except for stratification in DNR (3.8%). Using approach 2, residual confounding was present in CPRD with PS matching (5.3%) and stratification (6.4%), but not with weighting (4.3%). Within DNR, no NCOs had HR estimates with upper or lower CI limits beyond the specified bounds indicating residual confounding for any PS method. Achievement of covariate balance and determination of residual bias were dependent upon several factors including the population under study, PS method, prevalence of NCO, and the threshold indicating residual confounding. Lay Summary: Treatment groups in clinical practice may not be comparable as patient characteristics differ according to the need for the prescribed medication, known as confounding. We assessed comparability of 2 common osteoporosis treatments, denosumab and oral bisphosphonate, in 280 288 postmenopausal women using electronic health records from UK Clinical Practice Research Datalink (CPRD) and Danish National Registries (DNR). We evaluated comparability of recorded patient characteristics with 3 propensity score (PS) methods, matching, stratification, and weighting. We assessed residual confounding from unrecorded patient characteristics via negative control outcomes (NCOs), events known not to be associated with treatment such as delirium. We found that achieving comparability of osteoporosis treatment groups depended on the study population, PS method, and definition of residual confounding. Weighting and stratification performed the best in DNR and CPRD, respectively. Using a stricter threshold based on statistical significance for the NCO suggested that the treatment groups were not comparable, except for PS stratification in DNR. Applying clinically significant thresholds of treatment effect size showed comparability using weighting in CPRD and all PS methods in DNR. Studies should consider more than one PS method to test robustness and identify the largest number of NCO to give the greatest flexibility in detecting residual confounding. Graphical Abstract [ABSTRACT FROM AUTHOR]

Regression on observational data can fail to capture a causal relationship in the presence of unobserved confounding. Confounding strength measures this mismatch, but estimating it requires itself additional assumptions. A common assumption is the independence of causal mechanisms, which relies on concentration phenomena in high dimensions. While high dimensions enable the estimation of confounding strength, they also necessitate adapted estimators. In this paper, we derive the asymptotic behavior of the confounding strength estimator by Janzing and Schölkopf (2018) and show that it is generally not consistent. We then use tools from random matrix theory to derive an adapted, consistent estimator. [ABSTRACT FROM AUTHOR]

BACKGROUND: Adjusting for multiple biases usually involves adjusting for one bias at a time, with careful attention to the order in which these biases are adjusted. A novel, alternative approach to multiple-bias adjustment involves the simultaneous adjustment of all biases via imputation and/or regression weighting. The imputed value or weight corresponds to the probability of the missing data and serves to reconstruct the unbiased data that would be observed based on the provided assumptions of the degree of bias. METHODS: We motivate and describe the steps necessary to implement this method. We also demonstrate the validity of this method through a simulation study with an exposure-outcome relationship that is biased by uncontrolled confounding, exposure misclassification, and selection bias. RESULTS: The study revealed that a non-biased effect estimate can be obtained when correct bias parameters are applied. It also found that incorrect specification of every bias parameter by +/-25% still produced an effect estimate with less bias than the observed, biased effect. CONCLUSIONS: Simultaneous multi-bias analysis is a useful way of investigating and understanding how multiple sources of bias may affect naive effect estimates. This new method can be used to enhance the validity and transparency of real-world evidence obtained from observational, longitudinal studies.

Background: The change of morphokinetic pattern in aneuploid embryos will facilitate the non-invasive selection of euploid embryos. In this study, we investigated the impact of different chromosomal abnormalities on the morphokinetic patterns of embryonic development. Methods: Our cohort includes 939 time-lapse preimplantation genetic testing cycles performed between January 2019 and July 2022 at a single academic fertility center, with a total of 2876 biopsied blastocysts. Intracytoplasmic sperm injection, blastocyst culture, trophectoderm biopsy, time-lapse monitoring, and next-generation sequencing were performed. Results: After adjusting for patient- and cycle-related factors, six morphokinetic parameters (t5, P = 0.006; t8, P = 0.048; tSB, P

Epidemiologists spend a great deal of time on confounding in our teaching, in our methods development, and in our assessment of study results. This may give the impression that uncontrolled confounding is the biggest problem observational epidemiology faces, when in fact, other sources of bias such as selection bias, measurement error, missing data, and misalignment of zero time may often (especially if they are all present in a single study) lead to a stronger deviation from the truth. Compared with the amount of time we spend teaching how to address confounding in data analysis, we spend relatively little time teaching methods for simulating confounding (and other sources of bias) to learn their impact and develop plans to mitigate or quantify the bias. Here we review the accompanying paper by Desai et al (Am J Epidemiol. 2024;193(11):1600-1608), which uses simulation methods to quantify the impact of an unmeasured confounder when it is completely missing or when a proxy of the confounder is measured. We discuss how we can use simulations of sources of bias to ensure that we generate better and more valid study estimates, and we discuss the importance of simulating realistic datasets with plausible bias structures to guide data collection. This article is part of a Special Collection on Pharmacoepidemiology. [ABSTRACT FROM AUTHOR]

Abstract Background Cardiovascular disease (CVD) is the leading cause of death worldwide. It has been known for some considerable time that radiation is associated with excess risk of CVD. A recent systematic review of radiation and CVD highlighted substantial inter-study heterogeneity in effect, possibly a result of confounding or modifications of radiation effect by non-radiation factors, in particular by the major lifestyle/environmental/medical risk factors and latent period. Methods We assessed effects of confounding by lifestyle/environmental/medical risk factors on radiation-associated CVD and investigated evidence for modifying effects of these variables on CVD radiation dose–response, using data assembled for a recent systematic review. Results There are 43 epidemiologic studies which are informative on effects of adjustment for confounding or risk modifying factors on radiation-associated CVD. Of these 22 were studies of groups exposed to substantial doses of medical radiation for therapy or diagnosis. The remaining 21 studies were of groups exposed at much lower levels of dose and/or dose rate. Only four studies suggest substantial effects of adjustment for lifestyle/environmental/medical risk factors on radiation risk of CVD; however, there were also substantial uncertainties in the estimates in all of these studies. There are fewer suggestions of effects that modify the radiation dose response; only two studies, both at lower levels of dose, report the most serious level of modifying effect. Conclusions There are still large uncertainties about confounding factors or lifestyle/environmental/medical variables that may influence radiation-associated CVD, although indications are that there are not many studies in which there are substantial confounding effects of these risk factors.

AbstractThis article introduces a collection of four datasets, similar to Anscombe’s quartet, that aim to highlight the challenges involved when estimating causal effects. Each of the four datasets is generated based on a distinct causal mechanism: the first involves a collider, the second involves a confounder, the third involves a mediator, and the fourth involves the induction of M-Bias by an included factor. The article includes a mathematical summary of each dataset, as well as directed acyclic graphs that depict the relationships between the variables. Despite the fact that the statistical summaries and visualizations for each dataset are identical, the true causal effect differs, and estimating it correctly requires knowledge of the data-generating mechanism. These example datasets can help practitioners gain a better understanding of the assumptions underlying causal inference methods and emphasize the importance of gathering more information beyond what can be obtained from statistical tools alone. The article also includes R code for reproducing all figures and provides access to the datasets themselves through an R package named “quartets.” Supplementary materials for this article are available online.

With linear dependency between the explanatory variables, partial least squares (PLS) regression is commonly used for regression analysis. If the response variable correlates to a high degree with the explanatory variables, a model with excellent predictive ability can usually be obtained. Ranking of variable importance is commonly used to interpret the model and sometimes this interpretation guides further experimentation. For instance, when analyzing natural product extracts for bioactivity, an underlying assumption is that the highest ranked compounds represent the best candidates for isolation and further testing. A problem with this approach is that in most cases, the number of compounds is larger than the number of samples (and usually much larger) and that the concentrations of the compounds correlate. Furthermore, compounds may interact as synergists or as antagonists. If the modeling process does not account for this possibility, the interpretation can be thoroughly wrong because unmodeled variables that strongly influence the response will give rise to confounding of a first‐order PLS model and send the experimenter on a wrong track. We show the consequences of this by a practical example from natural product research. Furthermore, we show that by including the possibility of interactions between explanatory variables, visualization using a selectivity ratio plot may provide model interpretation that can be used to make inferences. [ABSTRACT FROM AUTHOR]

Studying causal effects of continuous treatments is important for gaining a deeper understanding of many interventions, policies, or medications, yet researchers are often left with observational studies for doing so. In the observational setting, confounding is a barrier to the estimation of causal effects. Weighting approaches seek to control for confounding by reweighting samples so that confounders are comparable across different treatment values. Yet, for continuous treatments, weighting methods are highly sensitive to model misspecification. In this article we elucidate the key property that makes weights effective in estimating causal quantities involving continuous treatments. We show that to eliminate confounding, weights should make treatment and confounders independent on the weighted scale. We develop a measure that characterizes the degree to which a set of weights induces such independence. Further, we propose a new model-free method for weight estimation by optimizing our measure. We study the theoretical properties of our measure and our weights, and prove that our weights can explicitly mitigate treatment-confounder dependence. The empirical effectiveness of our approach is demonstrated in a suite of challenging numerical experiments, where we find that our weights are quite robust and work well under a broad range of settings. for this article are available online. [ABSTRACT FROM AUTHOR]

This study summarizes a presentation at the symposium for the Calvin Schwabe Award for Lifetime Achievement in Veterinary Epidemiology and Preventive Medicine, which was awarded to the first author. As epidemiologists, we are taught that “correlation does not imply causation.” While true, identifying causes is a key objective for much of the research that we conduct. There is empirical evidence that veterinary epidemiologists are conducting observational research with the intent to identify causes; many studies include control for confounding variables, and causal language is often used when interpreting study results. Frameworks for studying causes include the articulation of specific hypotheses to be tested, approaches for the selection of variables, methods for statistical estimation of the relationship between the exposure and the outcome, and interpretation of that relationship as causal. When comparing observational studies in veterinary populations to those conducted in human populations, the application of each of these steps differs substantially. The a priori identification of exposure- outcome pairs of interest are less common in observational studies in the veterinary literature compared to the human literature, and prior knowledge is used to select confounding variables in most observational studies in human populations, whereas data-driven approaches are the norm in veterinary populations. The consequences of not having a defined exposure- outcome hypotheses of interest and using data-driven analytical approaches include an increased probability of biased results and poor replicability of results. A discussion by the community of researchers on current approaches to studying causes in observational studies in veterinary populations is warranted. [ABSTRACT FROM AUTHOR]

In the last two decades, significant research efforts have been dedicated to addressing the issue of spatial confounding in linear regression models. Confounding occurs when the relationship between the covariate and the response variable is influenced by an unmeasured confounder associated with both. This results in biased estimators for the regression coefficients reduced efficiency, and misleading interpretations. This article aims to understand how confounding relates to the parameters of the data generating process. The sampling properties of the regression coefficient estimator are derived as ratios of dependent quadratic forms in Gaussian random variables: this allows us to obtain exact expressions for the marginal bias and variance of the estimator, that were not obtained in previous studies. Moreover, we provide an approximate measure of the marginal bias that gives insights of the main determinants of bias. Applications in the framework of geostatistical and areal data modeling are presented. Particular attention is devoted to the difference between smoothness and variability of random vectors involved in the data generating process. Results indicate that marginal covariance between the covariate and the confounder, along with marginal variability of the covariate, play the most relevant role in determining the magnitude of confounding, as measured by the bias. [ABSTRACT FROM AUTHOR]

Social work aspires to change social outcomes through policy and practice. Thus, researchers often use quantitative analysis to understand how social phenomena or policy and practice interventions change outcomes, but these causal questions cannot be answered using statistical association only. Directed acyclic graphs (DAGs), also called causal graphs, are a framework for representing assumptions about the causal relations between variables, imbuing statistical associations with causal meaning. This visual language may be more accessible than a math-based approach. We provide social work researchers with clear guidelines for using DAGs. First, we synthesize the current DAG literature and show how many statistical phenomena are represented in DAGs. Second, we describe an ordered process for building a DAG around a causal question about the impact of a treatment on an outcome. Third, we introduce a structured process called "clock and grid" for specifying the confounders that must be accounted for to estimate the unbiased causal effect. Throughout, we use working examples of social work evaluation scenarios to facilitate understanding of these concepts. We argue that DAGs represent an end-to-end conceptual framework for curating subject area knowledge that can advance social work research by informing design and analysis. [ABSTRACT FROM AUTHOR]

Epidemiological researchers often examine associations between risk factors and health outcomes in non-experimental designs. Observed associations may be causal or confounded by unmeasured factors. Sibling and co-twin control studies account for familial confounding by comparing exposure levels among siblings (or twins). If the exposure-outcome association is causal, the siblings should also differ regarding the outcome. However, such studies may sometimes introduce more bias than they alleviate. Measurement error in the exposure may bias results and lead to erroneous conclusions that truly causal exposure-outcome associations are confounded by familial factors. The current study used Monte Carlo simulations to examine bias due to measurement error in sibling control models when the observed exposure-outcome association is truly causal. The results showed that decreasing exposure reliability and increasing sibling-correlations in the exposure led to deflated exposure-outcome associations and inflated associations between the family mean of the exposure and the outcome. The risk of falsely concluding that causal associations were confounded was high in many situations. For example, when exposure reliability was 0.7 and the observed sibling-correlation was r = 0.4, about 30–90% of the samples (n = 2,000) provided results supporting a false conclusion of confounding, depending on how p-values were interpreted as evidence for a family effect on the outcome. The current results have practical importance for epidemiological researchers conducting or reviewing sibling and co-twin control studies and may improve our understanding of observed associations between risk factors and health outcomes. We have developed an app (SibSim) providing simulations of many situations not presented in this paper. [ABSTRACT FROM AUTHOR]

This commentary addresses a series of articles in Prevention Science about environmental causes of attention deficit hyperactivity disorder (ADHD). It provides an overview of their key findings and places them in a broader context to facilitate their interpretation. Each of the articles included in the special issue is a meta-analysis assessing the association of ADHD with several environmental exposures. Each of the author teams systematically searched for articles and defined eligibility criteria. They assured that the measurement of risk factors preceded the measurement of ADHD. Most of the analyses are based on many studies with many participants in the constituent studies. As is typical of any observational epidemiologic study, the constituent studies could not correct for all possible confounds because some were not measured, and some are unknown. For this reason, these meta-analyses may have documented confounded associations, which calls for cautious interpretations. None of the constituent studies assessed what might be the most important type of confounding, familial, and genetic confounding, which occurs when the environmental exposure being studied is correlated with the genetic risk of the disorder being studied or other familial risk factors. Addressing familial/genetic confounding requires a genetically informed study. Because of these issues, the results presented here are intriguing but require further examination before one can conclude that the reported associations correspond to causal events for ADHD. A pathway forward is suggested by drawing parallels between genomic and exposure research. [ABSTRACT FROM AUTHOR]

This article introduces a collection of four datasets, similar to Anscombe's quartet, that aim to highlight the challenges involved when estimating causal effects. Each of the four datasets is generated based on a distinct causal mechanism: the first involves a collider, the second involves a confounder, the third involves a mediator, and the fourth involves the induction of M-Bias by an included factor. The article includes a mathematical summary of each dataset, as well as directed acyclic graphs that depict the relationships between the variables. Despite the fact that the statistical summaries and visualizations for each dataset are identical, the true causal effect differs, and estimating it correctly requires knowledge of the data-generating mechanism. These example datasets can help practitioners gain a better understanding of the assumptions underlying causal inference methods and emphasize the importance of gathering more information beyond what can be obtained from statistical tools alone. The article also includes R code for reproducing all figures and provides access to the datasets themselves through an R package named "quartets." for this article are available online. [ABSTRACT FROM AUTHOR]

The relationship between body mass index (BMI) and outcomes in the acute care setting is controversial, with evidence suggesting that obesity is either protective - which is also called obesity paradox - or associated with worse outcomes. The purpose of this study was to assess whether BMI was related to frailty and biological age, and whether BMI remained predictive of mortality after adjusting for frailty and biological age. Of the 2950 patients who had a biological age estimated on admission to the intensive care unit, 877 (30 %) also had BMI and frailty data available for further analysis in this retrospective cohort study. Biological age of each patient was estimated using the Levine PhenoAge model based on results of nine blood tests that were reflective of DNA methylation. Biological age in excess of chronological age was then indexed to the local study context by a linear regression to generate the residuals. The associations between BMI, clinical frailty scale, and the residuals were first analyzed using univariable analyses. Their associations with mortality were then assessed by multivariable analysis, including the use of a 3-knot restricted cubic spline function to allow non-linearity. Both frailty (p = 0.003) and the residuals of the biological age (p = 0.001) were related to BMI in a U-shaped fashion. BMI was not related to hospital mortality, but both frailty (p = 0.015) and the residuals of biological age (OR per decade older than chronological age 1.50, 95 % confidence interval [CI] 1.04–2.18; p = 0.031) were predictive of mortality after adjusting for chronological age, diabetes mellitus and severity of acute illness. BMI was significantly associated with both frailty and biological age in a U-shaped fashion but only the latter two were related to mortality. These results may, in part, explain why obesity paradox could be observed in some studies. [ABSTRACT FROM AUTHOR]

Controlling for confounding bias is crucial in causal inference. Causal inference using data from observational studies (e.g., electronic health records) or imperfectly randomized trials (e.g., imperfect randomization or compliance) requires accounting for confounding variables. Many different methods are currently employed to mitigate bias due to confounding. This paper provides a comprehensive review and tutorial of common estimands and confounding adjustment approaches, including outcome regression, g-computation, propensity score, and doubly robust methods. We discuss bias and precision, advantages and disadvantages, and software implementation for each method. Moreover, approaches are illustrated empirically with a reproducible case study. We conclude that different scientific questions are better addressed by certain estimands. No estimand is uniformly more appropriate. Upon selecting an estimand, decisions on which estimator can be driven by performance and available background knowledge. [ABSTRACT FROM AUTHOR]

Trial emulations in observational data analyses can complement findings from randomized clinical trials, inform future trial designs, or generate evidence when randomized studies are not feasible due to resource constraints and ethical or practical limitations. Importantly, trial emulation designs facilitate causal inference in observational data analyses by enhancing counterfactual thinking and comparisons of real-world observations (e.g. Mendelian Randomization) to hypothetical interventions. In order to enhance credibility, trial emulations would benefit from prospective registration, publication of statistical analysis plans, and subsequent prospective benchmarking to randomized clinical trials prior to their publication. Confounding by indication, however, is the key challenge to interpreting observed intended effects of an intervention as causal in observational data analyses. We discuss the target trial emulation of the REDUCE-AMI randomized clinical trial (ClinicalTrials.gov ID NCT03278509; beta-blocker use in patients with preserved left ventricular ejection fraction after myocardial infarction) to illustrate the challenges and uncertainties of studying intended effects of interventions without randomization to account for confounding. We furthermore directly compare the findings, statistical power, and clinical interpretation of the results of the REDUCE-AMI target trial emulation to those from the simultaneously published randomized clinical trial. The complexity and subtlety of confounding by indication when studying intended effects of interventions can generally only be addressed by randomization. [ABSTRACT FROM AUTHOR]

Excessive urbanization leads to considerable nature deficiency and abundant artificial infrastructure in urban areas, which triggered intensive discussions on people's exposure to green space and outdoor artificial light at night (ALAN). Recent academic progress highlights that people's exposure to green space and outdoor ALAN may be confounders of each other but lacks systematic investigations. This study investigates the associations between people's exposure to green space and outdoor ALAN by adopting the three most used research paradigms: population-level residence-based, individual-level residence-based, and individual-level mobility-oriented paradigms. We employed the green space and outdoor ALAN data of 291 Tertiary Planning Units in Hong Kong for population-level analysis. We also used data from 940 participants in six representative communities for individual-level analyses. Hong Kong green space and outdoor ALAN were derived from high-resolution remote sensing data. The total exposures were derived using the spatiotemporally weighted approaches. Our results confirm that the negative associations between people's exposure to green space and outdoor ALAN are universal across different research paradigms, spatially non-stationary, and consistent among different socio-demographic groups. We also observed that mobility-oriented measures may lead to stronger negative associations than residence-based measures by mitigating the contextual errors of residence-based measures. Our results highlight the potential confounding associations between people's exposure to green space and outdoor ALAN, and we strongly recommend relevant studies to consider both of them in modeling people's health outcomes, especially for those health outcomes impacted by the co-exposure to them.

Background: While prevalence estimates differ by definition of metabolic syndrome (MetS), it is less clear how different definitions affect associations with alcohol consumption. Methods: We included 3051 adults aged 25–77 from the baseline examination of the Swedish INTERGENE cohort (2001–2004). Using multiple logistic regression, we investigated cross-sectional associations between ethanol intake and MetS defined according to the Adult Treatment Panel III (ATP III), the International Diabetes Federation (IDF), and the Joint Interim Statement (JIS). Alcohol exposure categories comprised abstinence, and low, medium, and high consumption defined via sex-specific tertiles of ethanol intake among current consumers. Covariates included sociodemographics, health, and lifestyle factors. Results: MetS prevalence estimates varied between 13.9 % (ATP III) and 25.3 % (JIS), with higher prevalence in men than women. Adjusted for age and sex, medium-high alcohol consumption was associated with lower odds of MetS compared to low consumption, while no difference was observed for abstainers. Only the most specific (and thus severe) definition of MetS (ATP III) showed decreasing odds for ethanol intake when adjusted for all covariates. Conclusion: Our study shows that alcohol-related associations differ by definition of MetS. The finding that individuals with the most stringently defined MetS may benefit from alcohol consumption calls for further well-controlled studies.

Abstract Background Mammography screening programmes (MSP) aim to reduce breast cancer mortality by shifting diagnoses to earlier stages. However, it is difficult to evaluate the effectiveness of current MSP because analyses can only rely on observational data, comparing women who participate in screening with women who do not. These comparisons are subject to several biases: one of the most important is self-selection into the MSP, which introduces confounding and is difficult to control for. Here, we propose an approach to quantify confounding based on breast cancer survival analyses using readily available routine data sources. Methods Using data from the Cancer Registry of North Rhine-Westphalia, Germany, we estimate the relative contribution of confounding to the observed survival benefit of participants of the German MSP. This is accomplished by comparing non-participants, participants with screen-detected and participants with interval breast cancers for the endpoints “death from breast cancer” and “death from all causes other than breast cancer” — the latter being assumed to be unrelated to any MSP effect. By using different contrasts, we eliminate the effects of stage shift, lead and length time bias. The association of breast cancer detection mode with survival is analysed using Cox models in 68,230 women, aged 50–69 years, with breast cancer diagnosed in 2006–2014 and followed up until 2018. Results The hazard of dying from breast cancer was lower in participants with screen-detected cancer than in non-participants (HR = 0.21, 95% CI: 0.20–0.22), but biased by lead and length time bias, and confounding. When comparing participants with interval cancers and non-participants, the survival advantage was considerably smaller (HR = 0.62, 95% CI: 0.58–0.66), due to the elimination of stage shift and lead time bias. Finally, considering only mortality from causes other than breast cancer in the latter comparison, length time bias was minimised, but a survival advantage was still present (HR = 0.63, 95% CI: 0.56–0.70), which we attribute to confounding. Conclusions This study shows that, in addition to stage shift, lead and length time bias, confounding is an essential component when comparing the survival of MSP participants and non-participants. We further show that the confounding effect can be quantified without explicit knowledge of potential confounders by using a negative control outcome.

This article discusses a study conducted in Peru that aimed to assess the effectiveness of replacing sodium chloride with a salt substitute in reducing blood pressure and the incidence of hypertension in a community-wide setting. The study utilized a stepped-wedge cluster randomized trial design, where the intervention was introduced in a phased manner across different villages. The results showed that the salt substitute was associated with a modest reduction in blood pressure and a lower incidence of hypertension, particularly among individuals with hypertension at baseline. The study suggests that this intervention could be implemented on a population-wide basis. However, it is important to note that the study design used does not provide the same level of evidence as a randomized controlled trial. [Extracted from the article]

BackgroundEstimates of effect heterogeneity (i.e. the extent to which the causal effect of one exposure varies across strata of a second exposure) can be biased if the exposure-outcome relationship is subject to uncontrolled confounding whose severity differs across strata of the second exposure.MethodsWe propose methods, analogous to the E-value for total effects, that help to assess the sensitivity of effect heterogeneity estimates to possible uncontrolled confounding. These E-value analogues characterize the severity of uncontrolled confounding strengths that would be required, hypothetically, to 'explain away' an estimate of multiplicative or additive effect heterogeneity in the sense that appropriately controlling for those confounder(s) would have shifted the effect heterogeneity estimate to the null, or alternatively would have shifted its confidence interval to include the null. One can also consider shifting the estimate or confidence interval to an arbitrary non-null value. All of these E-values can be obtained using the R package EValue.ResultsWe illustrate applying the proposed E-value analogues to studies on: (i) effect heterogeneity by sex of the effect of educational attainment on dementia incidence and (ii) effect heterogeneity by age on the effect of obesity on all-cause mortality.ConclusionReporting these proposed E-values could help characterize the robustness of effect heterogeneity estimates to potential uncontrolled confounding.