Search

A common view in epidemiology is that automated confounder selection methods, such as backward elimination, should be avoided as they can lead to biased effect estimates and underestimation of their variance. Nevertheless, backward elimination remains regularly applied. We investigated if and under which conditions causal effect estimation in observational studies can improve by using backward elimination on a prespecified set of potential confounders. An expression was derived that quantifies how variable omission relates to bias and variance of effect estimators. Additionally, 3960 scenarios were defined and investigated by simulations comparing bias and mean squared error (MSE) of the conditional log odds ratio, log(cOR), and the marginal log risk ratio, log(mRR), between full models including all prespecified covariates and backward elimination of these covariates. Applying backward elimination resulted in a mean bias of 0.03 for log(cOR) and 0.02 for log(mRR), compared to 0.56 and 0.52 for log(cOR) and log(mRR), respectively, for a model without any covariate adjustment, and no bias for the full model. In less than 3% of the scenarios considered, the MSE of the log(cOR) or log(mRR) was slightly lower (max 3%) when backward elimination was used compared to the full model. When an initial set of potential confounders can be specified based on background knowledge, there is minimal added value of backward elimination. We advise not to use it and otherwise to provide ample arguments supporting its use., (© 2022 The Authors. Biometrical Journal published by Wiley-VCH GmbH.)

Purpose: In studies of effects of time-varying drug exposures, adequate adjustment for time-varying covariates is often necessary to properly control for confounding. However, the granularity of the available covariate data may not be sufficiently fine, for example when covariates are measured for participants only when their exposure levels change., Methods: To illustrate the impact of choices regarding the frequency of measuring time-varying covariates, we simulated data for a large target trial and for large observational studies, varying in covariate measurement design. Covariates were measured never, on a fixed-interval basis, or each time the exposure level switched. For the analysis, it was assumed that covariates remain constant in periods of no measurement. Cumulative survival probabilities for continuous exposure and non-exposure were estimated using inverse probability weighting to adjust for time-varying confounding, with special emphasis on the difference between 5-year event risks., Results: With monthly covariate measurements, estimates based on observational data coincided with trial-based estimates, with 5-year risk differences being zero. Without measurement of baseline or post-baseline covariates, this risk difference was estimated to be 49% based on the available observational data. With measurements on a fixed-interval basis only, 5-year risk differences deviated from the null, to 29% for 6-monthly measurements, and with magnitude increasing up to 35% as the interval length increased. Risk difference estimates diverged from the null to as low as -18% when covariates were measured depending on exposure level switching., Conclusion: Our simulations highlight the need for careful consideration of time-varying covariates in designing studies on time-varying exposures. We caution against implementing designs with long intervals between measurements. The maximum length required will depend on the rates at which treatments and covariates change, with higher rates requiring shorter measurement intervals., (© 2021 The Authors. Pharmacoepidemiology and Drug Safety published by John Wiley & Sons Ltd.)

Statistical correction for measurement error in epidemiologic studies is possible, provided that information about the measurement error model and its parameters are available. Such information is commonly obtained from a randomly sampled internal validation sample. It is however unknown whether randomly sampling the internal validation sample is the optimal sampling strategy. We conducted a simulation study to investigate various internal validation sampling strategies in conjunction with regression calibration. Our simulation study showed that for an internal validation study sample of 40% of the main study's sample size, stratified random and extremes sampling had a small efficiency gain over random sampling (10% and 12% decrease on average over all scenarios, respectively). The efficiency gain was more pronounced in smaller validation samples of 10% of the main study's sample size (i.e., a 31% and 36% decrease on average over all scenarios, for stratified random and extremes sampling, respectively). To mitigate the bias due to measurement error in epidemiologic studies, small efficiency gains can be achieved for internal validation sampling strategies other than random, but only when measurement error is nondifferential. For regression calibration, the gain in efficiency is, however, at the cost of a higher percentage bias and lower coverage., (© The Author(s) 2021. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health.)

Observational data are increasingly used with the aim of estimating causal effects of treatments, through careful control for confounding. Marginal structural models estimated using inverse probability weighting (MSMs-IPW), like other methods to control for confounding, assume that confounding variables are measured without error. The average treatment effect in an MSM-IPW may however be biased when a confounding variable is error prone. Using the potential outcome framework, we derive expressions for the bias due to confounder misclassification in analyses that aim to estimate the average treatment effect using an marginal structural model estimated using inverse probability weighting (MSM-IPW). We compare this bias with the bias due to confounder misclassification in analyses based on a conditional regression model. Focus is on a point-treatment study with a continuous outcome. Compared with bias in the average treatment effect in a conditional model, the bias in an MSM-IPW can be different in magnitude but is equal in sign. Also, we use a simulation study to investigate the finite sample performance of MSM-IPW and conditional models when a confounding variable is misclassified. Simulation results indicate that confidence intervals of the treatment effect obtained from MSM-IPW are generally wider, and coverage of the true treatment effect is higher compared with a conditional model, ranging from overcoverage if there is no confounder misclassification to undercoverage when there is confounder misclassification. Further, we illustrate in a study of blood pressure-lowering therapy, how the bias expressions can be used to inform a quantitative bias analysis to study the impact of confounder misclassification, supported by an online tool.

The validity of clinical research is potentially threatened by missing data. Any variable measured in a study can have missing values, including the exposure, the outcome, and confounders. When missing values are ignored in the analysis, only those subjects with complete records will be included in the analysis. This may lead to biased results and loss of power. We explain why missing data may lead to bias and discuss a commonly used classification of missing data.

Epidemiologists are often confronted with datasets to analyse which contain measurement error due to, for instance, mistaken data entries, inaccurate recordings and measurement instrument or procedural errors. If the effect of measurement error is misjudged, the data analyses are hampered and the validity of the study's inferences may be affected. In this paper, we describe five myths that contribute to misjudgments about measurement error, regarding expected structure, impact and solutions to mitigate the problems resulting from mismeasurements. The aim is to clarify these measurement error misconceptions. We show that the influence of measurement error in an epidemiological data analysis can play out in ways that go beyond simple heuristics, such as heuristics about whether or not to expect attenuation of the effect estimates. Whereas we encourage epidemiologists to deliberate about the structure and potential impact of measurement error in their analyses, we also recommend exercising restraint when making claims about the magnitude or even direction of effect of measurement error if not accompanied by statistical measurement error corrections or quantitative bias analysis. Suggestions for alleviating the problems or investigating the structure and magnitude of measurement error are given., (© The Author(s) 2019. Published by Oxford University Press on behalf of the International Epidemiological Association.)

Objectives: In medical research, covariates (e.g., exposure and confounder variables) are often measured with error. While it is well accepted that this introduces bias and imprecision in exposure-outcome relations, it is unclear to what extent such issues are currently considered in research practice. The objective was to study common practices regarding covariate measurement error via a systematic review of general medicine and epidemiology literature., Study Design and Setting: Original research published in 2016 in 12 high impact journals was full-text searched for phrases relating to measurement error. Reporting of measurement error and methods to investigate or correct for it were quantified and characterized., Results: Two hundred and forty-seven (44%) of the 565 original research publications reported on the presence of measurement error. 83% of these 247 did so with respect to the exposure and/or confounder variables. Only 18 publications (7% of 247) used methods to investigate or correct for measurement error., Conclusions: Consequently, it is difficult for readers to judge the robustness of presented results to the existence of measurement error in the majority of publications in high impact journals. Our systematic review highlights the need for increased awareness about the possible impact of covariate measurement error. Additionally, guidance on the use of measurement error correction methods is necessary., (Copyright © 2018 Elsevier Inc. All rights reserved.)

Observational drug safety studies may be susceptible to confounding or protopathic bias. This bias may cause a spurious relationship between drug exposure and adverse side effect when none exists and may lead to unwarranted safety alerts. The spurious relationship may manifest itself through substantially different risk levels between exposure groups at the start of follow-up when exposure is deemed too short to have any plausible biological effect of the drug. The restrictive proportional hazards assumption with its arbitrary choice of baseline hazard function renders the commonly used Cox proportional hazards model of limited use for revealing such potential bias. We demonstrate a fully parametric approach using accelerated failure time models with an illustrative safety study of glucose-lowering therapies and show that its results are comparable against other methods that allow time-varying exposure effects. Our approach includes a wide variety of models that are based on the flexible generalized gamma distribution and allows direct comparisons of estimated hazard functions following different exposure-specific distributions of survival times. This approach lends itself to two alternative metrics, namely relative times and difference in times to event, allowing physicians more ways to communicate patient's prognosis without invoking the concept of risks, which some may find hard to grasp. In our illustrative case study, substantial differences in cancer risks at drug initiation followed by a gradual reduction towards null were found. This evidence is compatible with the presence of protopathic bias, in which undiagnosed symptoms of cancer lead to switches in diabetes medication., (Copyright © 2015 John Wiley & Sons, Ltd.)

Purpose: To assess the impact of random left truncation of data on the estimation of time-dependent exposure effects., Methods: A simulation study was conducted in which the relation between exposure and outcome was based on an immediate exposure effect, a first-time exposure effect, or a cumulative exposure effect. The individual probability of truncation, the moment of truncation, the exposure rate, and the incidence rate of the outcome were varied in different simulations. All observations before the moment of left truncation were omitted from the analysis., Results: Random left truncation did not bias estimates of immediate exposure effects, but resulted in an overestimation of a cumulative exposure effect and underestimation of a first-time exposure effect. The magnitude of bias in estimation of cumulative exposure effects depends on a combination of exposure rate, probability of truncation, and proportion of follow-up time left truncated., Conclusions: In case of a cumulative or first-time exposure, left truncation can result in substantial bias in pharmacoepidemiologic studies. The potential for this bias likely differs between databases, which may lead to heterogeneity in estimated exposure effects between studies., (Copyright © 2015 Elsevier Inc. All rights reserved.)

Background: Meta-analyses are typically triggered by a (potentially false-significant) finding in one of the preceding primary studies. We studied consequences of meta-analysis investigating effects when primary studies that triggered such meta-analysis are also included., Methods: We analytically determined the bias of the treatment effect estimates obtained by meta-analysis, conditional on the number of included primary and false-significant studies. The type I error rate and power of the meta-analysis were assessed using simulations. We applied a method for bias-correction, by subtracting an analytically derived bias from the treatment effect estimated in meta-analysis., Results: Bias in meta-analytical effects and type I error rates increased when increasing numbers of primary studies with false-significant effects were included. When 20% of the primary studies showed false-significant effects, the bias was 0.33 (z-score) instead of 0, and the type I error rate was 23% instead of 5%. After applying a bias-correction, the type I error rate became indeed 5%., Conclusions: Inclusion of primary studies with false-significant effects leads to biased effect estimates and inflated type I error rates in the meta-analysis, depending on the number of false-significant studies. This bias can be adjusted for.

A systematic distortion of the relationship between a treatment, risk factor or exposure and clinical outcomes is denoted by the term 'bias'. Three types of bias can be distinguished: information bias, selection bias, and confounding. These three types of bias and their potential solutions are discussed using various examples.

Confounding is an important and common issue in observational (non-randomized) research on the effects of pharmaceuticals or exposure to etiologic factors (determinants). Confounding is present when a third factor, related to both the determinant and the outcome, distorts the causal relation between these two. There are different methods to control for confounding. The most commonly used are restriction, stratification, multivariable regression models, and propensity score methods. With these methods it is only possible to control for variables for which data is known: measured confounders. Research in the area of confounding is currently directed at the incorporation of external knowledge on unmeasured confounders, the evaluation of instrumental variables, and the impact of time-dependent confounding.

Purpose: To systematically review observational studies on medical interventions to determine the quality of reporting of confounding., Methods: Articles on observational studies on medical interventions in five general medical journals and five epidemiological journals published between January 2004 and April 2007 were systematically reviewed. All relevant items pertaining to confounding bias were scored for each article. The overall quality of reporting was determined with an 8-point score., Results: The MEDLINE search resulted in 2993 publications, and 174 (5.8%) articles were included in the analysis. In the majority of studies (>98%), the potential for confounding bias was reported. Details on the selection and inclusion of observed confounders were reported in 10% and 51%, respectively. The potential for unobserved confounding was reported in 60%, and 9% commented on the potential effect of such remaining confounding. The quality of reporting of confounding score was mediocre (a median score of 4 points; interquartile range 3 to 5), and scores were similar in all years., Conclusion: The quality of reporting of confounding in articles on observational medical intervention studies was poor. However, the STROBE statement for reporting of observational studies may considerably impact the reporting of such studies.

We conducted a systematic literature search in Medline to assess the proportion of observational intervention studies appreciating confounding bias in peer-reviewed medical literature from 1985 through 2005. This study shows only 9% of all papers on observational intervention studies published in peer-reviewed medical journals mention any of the terms (confounding, adjustment, or bias) indicating appreciation of confounding.

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]

Directed acyclic graphs (DAGs), or causal diagrams, are graphical representations of causal structures that can be used in medical research to understand and illustrate potential bias, including bias arising from confounding, selection, and misclassification. Further, they provide guidance for researchers about how to address a potential bias. [ABSTRACT FROM AUTHOR]

For many statistical methods that are commonly used in medical research, it is assumed that observations are independent. However, when this assumption is violated, alternative methods may be needed. In this paper, we describe why an analysis that ignores the dependencies within the data may provide biased results and outline possible alternatives. This paper aims to create awareness among researchers, readers, and reviewers for this possible source of bias in medical research. [ABSTRACT FROM AUTHOR]

Purpose: During the first waves of the coronavirus pandemic, evidence on potential effective treatments was urgently needed. Results from observational studies on the effectiveness of hydroxychloroquine (HCQ) were conflicting, potentially due to biases. We aimed to assess the quality of observational studies on HCQ and its relation to effect sizes. Methods: PubMed was searched on 15 March 2021 for observational studies on the effectiveness of in‐hospital use of HCQ in COVID‐19 patients, published between 01/01/2020 and 01/03/2021 on. Study quality was assessed using the ROBINS‐I tool. Association between study quality and study characteristics (journal ranking, publication date, and time between submission and publication) and differences between effects sizes found in observational studies compared to those found in RCTs, were assessed using Spearman's correlation. Results: Eighteen of the 33 (55%) included observational studies were scored as critical risk of bias, eleven (33%) as serious risk and only four (12%) as moderate risk of bias. Biases were most often scored as critical in the domains related to selection of participants (n = 13, 39%) and bias due to confounding (n = 8, 24%). There were no significant associations found between the study quality and the characteristics nor between the study quality and the effect estimates. Discussion: Overall, the quality of observational HCQ studies was heterogeneous. Synthesis of evidence of effectiveness of HCQ in COVID‐19 should focus on RCTs and carefully consider the added value and quality of observational evidence. [ABSTRACT FROM AUTHOR]

In this methodology editorial, the problem of censored survival (or time-to-event) data in clinical research is discussed. In case of censored observations, part of the observation time for some participants in a study is unobserved, which means that measures such as an average survival time cannot be calculated. Instead, censored survival time should explicitly be accounted for in the analysis of a study, for example using the Kaplan–Meier method. Implications of censoring and how to deal with it in the analysis of a study or when reading a research paper are discussed. [ABSTRACT FROM AUTHOR]

Aim: Pragmatic clinical trials (PCTs) are randomized trials implemented through routine clinical practice, where design parameters of traditional randomized controlled trials are modified to increase generalizability. However, this may introduce statistical challenges. We aimed to identify these challenges and discuss possible solutions leading to best practice recommendations for the design and analysis of PCTs. Methods: A modified Delphi method was used to reach consensus among a panel of 11 experts in clinical trials and statistics. Statistical issues were identified in a focused literature review and aggregated with insights and possible solutions from experts collected through a series of survey iterations. Issues were ranked according to their importance. Results: Twenty‐seven articles were included and combined with experts' insight to generate a list of issues categorized into participants, recruiting sites, randomization, blinding and intervention, outcome (selection and measurement) and data analysis. Consensus was reached about the most important issues: risk of participants' attrition, heterogeneity of "usual care" across sites, absence of blinding, use of a subjective endpoint and data analysis aligned with the trial estimand. Potential issues should be anticipated and preferably be addressed in the trial protocol. The experts provided solutions regarding data collection and data analysis, which were considered of equal importance. Discussion: A set of important statistical issues in PCTs was identified and approaches were suggested to anticipate and/or minimize these through data analysis. Any impact of choosing a pragmatic design feature should be gauged in the light of the trial estimand. [ABSTRACT FROM AUTHOR]

Advice regarding the analysis of observational studies of exposure effects usually is against adjustment for factors that occur after the exposure, as they may be caused by the exposure (or mediate the effect of exposure on outcome), so potentially leading to collider stratification bias. However, such factors could also be caused by unmeasured confounding factors, in which case adjusting for them will also remove some of the bias due to confounding. We derive expressions for collider stratification bias when conditioning and confounding bias when not conditioning on the mediator, in the presence of unmeasured confounding (assuming that all associations are linear and there are no interactions). Using simulations, we show that generally neither the conditioned nor the unconditioned estimate is unbiased, and the trade-off between them depends on the magnitude of the effect of the exposure that is mediated relative to the effect of the unmeasured confounders and their relations with the mediator. We illustrate the use of the bias expressions via three examples: neuroticism and mortality (adjusting for the mediator appears the least biased option), glycated hemoglobin levels and systolic blood pressure (adjusting gives smaller bias), and literacy in primary school pupils (not adjusting gives smaller bias). Our formulae and simulations can inform quantitative bias analysis as well as analysis strategies for observational studies in which there is a potential for unmeasured confounding. [ABSTRACT FROM AUTHOR]

Epidemiologists are often confronted with datasets to analyse which contain measurement error due to, for instance, mistaken data entries, inaccurate recordings and measurement instrument or procedural errors. If the effect of measurement error is misjudged, the data analyses are hampered and the validity of the study's inferences may be affected. In this paper, we describe five myths that contribute to misjudgments about measurement error, regarding expected structure, impact and solutions to mitigate the problems resulting from mismeasurements. The aim is to clarify these measurement error misconceptions. We show that the influence of measurement error in an epidemiological data analysis can play out in ways that go beyond simple heuristics, such as heuristics about whether or not to expect attenuation of the effect estimates. Whereas we encourage epidemiologists to deliberate about the structure and potential impact of measurement error in their analyses, we also recommend exercising restraint when making claims about the magnitude or even direction of effect of measurement error if not accompanied by statistical measurement error corrections or quantitative bias analysis. Suggestions for alleviating the problems or investigating the structure and magnitude of measurement error are given. [ABSTRACT FROM AUTHOR]

Background The impact of unmeasured confounders on causal associations can be studied by means of sensitivity analyses. Although several sensitivity analyses are available, these are used infrequently. This article is intended as a tutorial on sensitivity analyses, in which we discuss three methods to conduct sensitivity analysis. [ABSTRACT FROM PUBLISHER]

Objectives: Unobserved confounding has been suggested to explain the effect of influenza vaccination on mortality reported in several observational studies. An instrumental variable (IV) is strongly related to the exposure under study, but not directly or indirectly (through other variables) with the outcome. Theoretically, analyses using IVs to control for both observed and unobserved confounding may provide unbiased estimates of influenza vaccine effects. We assessed the usefulness of IV analysis in influenza vaccination studies. Methods: Information on patients aged 65 years and older from the computerized Utrecht General Practitioner (GP) research database over seven influenza epidemic periods was pooled to estimate the association between influenza vaccination and all-cause mortality among community-dwelling elderly. Potential IVs included in the analysis were a history of gout, a history of orthopaedic morbidity, a history of antacid medication use, and GP-specific vaccination rates. Results: Using linear regression analyses, all possible IVs were associated with vaccination status: risk difference (RD) 7.8% (95% confidence interval [CI] 3.6%; 12.0%), RD 2.8% (95% CI 1.7%; 3.9%), RD 8.1% (95% CI 6.1%; 10.1%), and RD 100.0% (95% CI 89.0%; 111.0%) for gout, orthopaedic morbidity, antacid medication use, and GP-specific vaccination rates, respectively. Each potential IV, however, also appeared to be related to mortality through other observed confounding variables (notably age, sex, and comorbidity). Conclusions: The potential IVs studied did not meet the necessary criteria, because they were (indirectly) associated with the outcome. These variables may, therefore, not be suited to assess unconfounded influenza vaccine effects through IV analysis. [ABSTRACT FROM AUTHOR]

Background: Observational studies of medical interventions or risk factors are potentially biased by unmeasured confounding. In this paper we propose a Bayesian approach by defining an informative prior for the confounder-outcome relation, to reduce bias due to unmeasured confounding. This approach was motivated by the phenomenon that the presence of unmeasured confounding may be reflected in observed confounder-outcome relations being unexpected in terms of direction or magnitude.Methods: The approach was tested using simulation studies and was illustrated in an empirical example of the relation between LDL cholesterol levels and systolic blood pressure. In simulated data, a comparison of the estimated exposure-outcome relation was made between two frequentist multivariable linear regression models and three Bayesian multivariable linear regression models, which varied in the precision of the prior distributions. Simulated data contained information on a continuous exposure, a continuous outcome, and two continuous confounders (one considered measured one unmeasured), under various scenarios.Results: In various scenarios the proposed Bayesian analysis with an correctly specified informative prior for the confounder-outcome relation substantially reduced bias due to unmeasured confounding and was less biased than the frequentist model with covariate adjustment for one of the two confounding variables. Also, in general the MSE was smaller for the Bayesian model with informative prior, compared to the other models.Conclusions: As incorporating (informative) prior information for the confounder-outcome relation may reduce the bias due to unmeasured confounding, we consider this approach one of many possible sensitivity analyses of unmeasured confounding. [ABSTRACT FROM AUTHOR]