Search

Determining if the frequency distribution of a given data set follows a normal distribution or not is among the first steps of data analysis. Visual examination of the data, commonly by Q-Q plot, although is acceptable by many scientists, is considered subjective and not acceptable by other researchers. One-sample Kolmogorov-Smirnov test with Lilliefors correction (for a sample size ≥ 50) and Shapiro-Wilk test (for a sample size < 50) are common statistical tests for checking the normality of a data set quantitatively. As parametric tests, which assume that the data distribution is normal (Gaussian, bell-shaped), are more robust compared to their non-parametric counterparts, we commonly use transformations (e.g., log-transformation, Box-Cox transformation, etc.) to make the frequency distribution of non-normally distributed data close to a normal distribution. Herein, I wish to reflect on presenting how to practically work with these statistical methods through examining of real data sets., Competing Interests: The author has no potential conflicts of interest to disclose., (© 2024 The Korean Academy of Medical Sciences.)

Interpreting the results of epidemiologic studies calls for objectivity and rigorous scrutiny, acknowledging the limitations that temper the applicability of the findings to public health action. Current trends have posed new challenges to balancing goals of scientific objectivity and validity with public health applications. The ongoing tension between epidemiology's aspirations and capability has several sources: the need to overpromise in research proposals, compromising methodological rigor because of public health importance, defending findings in the face of hostile critics, and appealing to core constituencies who have specific expectations from the research., (© The Author(s) 2020. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.)

When estimating causal effects, careful handling of missing data is needed to avoid bias. Complete-case analysis is commonly used in epidemiologic analyses. Previous work has shown that covariate-stratified effect estimates from complete-case analysis are unbiased when missingness is independent of the outcome conditional on the exposure and covariates. Here, we assess the bias of complete-case analysis for adjusted marginal effects when confounding is present under various causal structures of missing data. We show that estimation of the marginal risk difference requires an unbiased estimate of the unconditional joint distribution of confounders and any other covariates required for conditional independence of missingness and outcome. The dependence of missing data on these covariates must be considered to obtain a valid estimate of the covariate distribution. If none of these covariates are effect-measure modifiers on the absolute scale, however, the marginal risk difference will equal the stratified risk differences and the complete-case analysis will be unbiased when the stratified effect estimates are unbiased. Estimation of unbiased marginal effects in complete-case analysis therefore requires close consideration of causal structure and effect-measure modification., (© The Author(s) 2020. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.)

Objective: Epidemiologic and clinical research papers often describe the study sample in the first table. If well-executed, this "Table 1" can illuminate potential threats to internal and external validity. However, little guidance exists on best practices for designing a Table 1, especially for complex study designs and analyses. We aimed to summarize and extend the literature related to reporting descriptive statistics., Study Design and Setting: In consultation with existing guidelines, we synthesized and developed reporting recommendations driven by study design and focused on transparency related to potential threats to internal and external validity., Results: We describe a basic structure for Table 1 and discuss simple modifications in terms of columns, rows, and cells to enhance a reader's ability to judge both internal and external validity. We further highlight several analytic complexities common in epidemiologic research (missing data, sample weights, clustered data, and interaction) and describe possible variations to Table 1 to maintain and add clarity about study validity in light of these issues. We discuss considerations and tradeoffs in Table 1 related to breadth and comprehensiveness vs. parsimony and reader-friendliness., Conclusion: We anticipate that our work will guide authors considering layouts for Table 1, with attention to the reader's perspective., (Copyright © 2019 Elsevier Inc. All rights reserved.)

Epidemiologists are attempting to address research questions of increasing complexity by developing novel methods for combining information from diverse sources. Cole et al. (Am J Epidemiol. 2023;192(3)467–474) provide 2 examples of the process of combining information to draw inferences about a population proportion. In this commentary, we consider combining information to learn about a target population as an epidemiologic activity and distinguish it from more conventional meta-analyses. We examine possible rationales for combining information and discuss broad methodological considerations, with an emphasis on study design, assumptions, and sources of uncertainty. [ABSTRACT FROM AUTHOR]

Propensity score analysis is a common approach to addressing confounding in nonrandomized studies. Its implementation, however, requires important assumptions (e.g. positivity). The disease risk score (DRS) is an alternative confounding score that can relax some of these assumptions. Like the propensity score, the DRS summarizes multiple confounders into a single score, on which conditioning by matching allows the estimation of causal effects. However, matching relies on arbitrary choices for pruning out data (e.g. matching ratio, algorithm, and caliper width) and may be computationally demanding. Alternatively, weighting methods, common in propensity score analysis, are easy to implement and may entail fewer choices, yet none have been developed for the DRS. Here we present 2 weighting approaches: One derives directly from inverse probability weighting; the other, named target distribution weighting , relates to importance sampling. We empirically show that inverse probability weighting and target distribution weighting display performance comparable to matching techniques in terms of bias but outperform them in terms of efficiency (mean squared error) and computational speed (up to >870 times faster in an illustrative study). We illustrate implementation of the methods in 2 case studies where we investigate placebo treatments for multiple sclerosis and administration of aspirin in stroke patients. [ABSTRACT FROM AUTHOR]

The aim of this study was to update and validate the Pregnancy Physical Activity Questionnaire (PPAQ), using novel and innovative accelerometer and wearable camera measures in a free-living setting, to improve the measurement performance of this method for self-reporting physical activity. A prospective cohort of 50 eligible pregnant women were enrolled in early pregnancy (mean = 14.9 weeks' gestation). In early, middle, and late pregnancy, participants completed the updated PPAQ and, for 7 days, wore an accelerometer (GT3X-BT; ActiGraph, Pensacola, Florida) on the nondominant wrist and a wearable camera (Autographer; OMG Life (defunct)). At the end of the 7-day period, participants repeated the PPAQ. Spearman correlations between the PPAQ and accelerometer data ranged from 0.37 to 0.44 for total activity, 0.17 to 0.53 for moderate- to vigorous-intensity activity, 0.19 to 0.42 for light-intensity activity, and 0.23 to 0.45 for sedentary behavior. Spearman correlations between the PPAQ and wearable camera data ranged from 0.52 to 0.70 for sports/exercise and from 0.26 to 0.30 for transportation activity. Reproducibility scores ranged from 0.70 to 0.92 for moderate- to vigorous-intensity activity and from 0.79 to 0.91 for sports/exercise, and were comparable across other domains of physical activity. The PPAQ is a reliable instrument and a valid measure of a broad range of physical activities during pregnancy. [ABSTRACT FROM AUTHOR]

Unlike parametric regression, machine learning (ML) methods do not generally require precise knowledge of the true data-generating mechanisms. As such, numerous authors have advocated for ML methods to estimate causal effects. Unfortunately, ML algorithms can perform worse than parametric regression. We demonstrate the performance of ML-based singly and doubly robust estimators. We used 100 Monte Carlo samples with sample sizes of 200, 1,200, and 5,000 to investigate bias and confidence-interval coverage under several scenarios. In a simple confounding scenario, confounders were related to the treatment and the outcome via parametric models. In a complex confounding scenario, the simple confounders were transformed to induce complicated nonlinear relationships. In the simple scenario, when ML algorithms were used, double-robust estimators were superior to singly robust estimators. In the complex scenario, single-robust estimators with ML algorithms were at least as biased as estimators using misspecified parametric models. Doubly robust estimators were less biased, but coverage was well below nominal. The use of sample splitting, inclusion of confounder interactions, reliance on a richly specified ML algorithm, and use of doubly robust estimators was the only explored approach that yielded negligible bias and nominal coverage. Our results suggest that ML-based singly robust methods should be avoided. [ABSTRACT FROM AUTHOR]

Objectives: Persons with Parkinson's disease (PD) have a higher prevalence of oral diseases and orofacial dysfunction, but knowledge about the use of dental care and whether their dental care needs are met is sparse. This study aimed to investigate the dental attendance and usage of dental care services of the total PD population in Denmark and compare it with a control group. Methods: National registers were used to identify the total PD population in Denmark (n = 6874) and to obtain data on their dental care from 2015 to 2019. These data were compared with a five‐fold age‐, gender‐ and geographically matched control group without PD (n = 34 285). Register data on age, gender, civil status, educational level, income, nursing homes status and mortality were also collected and adjusted for in the analyses. The dental attendance was analysed using χ2‐test with Bonferroni correction, and the type of dental care services was analysed using negative binomial regression analysis. Results: A significantly higher proportion of persons with PD were irregular attenders of the dental care system (21.0%), compared with the control group (16.9%). Persons with PD had a significantly higher overall usage of dental cares services. Most prominent was the high usage of treatment services, where persons with PD had a 1.50 times higher incidence rate of tooth extractions and a 1.71 times higher incidence rate of tooth fillings in the five years compared with the control group. Conclusion: Persons with PD are more often irregular users of dental care and receive more treatment services than the control group. This indicates a need for high‐quality prophylactic initiatives to prevent high filling and tooth extraction rates. Furthermore, this knowledge can be used by clinicians and decision makers to ensure optimal dental care for persons with PD. [ABSTRACT FROM AUTHOR]

Simulation methods are a powerful set of tools that can allow researchers to better characterize phenomena from the real world. As such, the ability to simulate data represents a critical set of skills that epidemiologists should use to better understand epidemiologic concepts and ensure that they have the tools to continue to self-teach even when their formal instruction ends. Simulation methods are not always taught in epidemiology methods courses, whereas causal directed acyclic graphs (DAGs) often are. Therefore, this paper details an approach to building simulations from DAGs and provides examples and code for learning to perform simulations. We recommend using very simple DAGs to learn the procedures and code necessary to set up a simulation that builds on key concepts frequently of interest to epidemiologists (e.g. mediation, confounding bias, M bias). We believe that following this approach will allow epidemiologists to gain confidence with a critical skill set that may in turn have a positive impact on how they conduct future epidemiologic studies. [ABSTRACT FROM AUTHOR]

In this paper, we consider methods for generating draws of a binary random variable whose expectation conditional on covariates follows a logistic regression model with known covariate coefficients. We examine approximations for finding a "balancing intercept," that is, a value for the intercept of the logistic model that leads to a desired marginal expectation for the binary random variable. We show that a recently proposed analytical approximation can produce inaccurate results, especially when targeting more extreme marginal expectations or when the linear predictor of the regression model has high variance. We then formulate the balancing intercept as a solution to an integral equation, implement a numerical approximation for solving the equation based on Monte Carlo methods, and show that the approximation works well in practice. Our approach to the basic problem of the balancing intercept provides an example of a broadly applicable strategy for formulating and solving problems that arise in the design of simulation studies used to evaluate or teach epidemiologic methods. [ABSTRACT FROM AUTHOR]

Data-sharing improves epidemiologic research, but the sharing of data frustrates epidemiologic researchers. The inefficiencies of current methods and options for data-sharing are increasingly documented and easily understood by any study group that has shared its data and any researcher who has received shared data. In this issue of the Journal , Temprosa et al. (Am J Epidemiol. 2021;191(1):147–158) describe how the Consortium of Metabolomics Studies (COMETS) developed and deployed a flexible analytical platform to eliminate key pain points in large-scale metabolomics research. COMETS Analytics includes an online tool, but its cloud computing and technology are the supporting rather than the leading actors in this script. The COMETS team identified the need to standardize diverse and inconsistent metabolomics and covariate data and models across its many participating cohort studies, and then developed a flexible tool that gave its member studies choices about how they wanted to meet the consortium's analytical requirements. Different specialties will have different specific research needs and will probably continue to use and develop an array of diverse analytical and technical solutions for their projects. COMETS Analytics shows how important—and enabling—the upstream attention to data standards and data consistency is to producing high-quality metabolomics, consortia-based, and large-scale epidemiology research. [ABSTRACT FROM AUTHOR]

An increasing number of recent studies have suggested that doubly robust estimators with cross-fitting should be used when estimating causal effects with machine learning methods. However, not all existing programs that implement doubly robust estimators support machine learning methods and cross-fitting, or provide estimates on multiplicative scales. To address these needs, we developed AIPW , a software package implementing augmented inverse probability weighting (AIPW) estimation of average causal effects in R (R Foundation for Statistical Computing, Vienna, Austria). Key features of the AIPW package include cross-fitting and flexible covariate adjustment for observational studies and randomized controlled trials (RCTs). In this paper, we use a simulated RCT to illustrate implementation of the AIPW estimator. We also perform a simulation study to evaluate the performance of the AIPW package compared with other doubly robust implementations, including CausalGAM , npcausal , tmle , and tmle3. Our simulation showed that the AIPW package yields performance comparable to that of other programs. Furthermore, we also found that cross-fitting substantively decreases the bias and improves the confidence interval coverage for doubly robust estimators fitted with machine learning algorithms. Our findings suggest that the AIPW package can be a useful tool for estimating average causal effects with machine learning methods in RCTs and observational studies. [ABSTRACT FROM AUTHOR]

To extend previous simulations on the performance of propensity score (PS) weighting and trimming methods to settings without and with unmeasured confounding, Poisson outcomes, and various strengths of treatment prediction (PS c statistic), we simulated studies with a binary intended treatment T as a function of 4 measured covariates. We mimicked treatment withheld and last-resort treatment by adding 2 "unmeasured" dichotomous factors that directed treatment to change for some patients in both tails of the PS distribution. The number of outcomes Y was simulated as a Poisson function of T and confounders. We estimated the PS as a function of measured covariates and trimmed the tails of the PS distribution using 3 strategies ("Crump," "Stürmer," and "Walker"). After trimming and reestimation, we used alternative PS weights to estimate the treatment effect (rate ratio): inverse probability of treatment weighting, standardized mortality ratio (SMR)-treated, SMR-untreated, the average treatment effect in the overlap population (ATO), matching, and entropy. With no unmeasured confounding, the ATO (123%) and "Crump" trimming (112%) improved relative efficiency compared with untrimmed inverse probability of treatment weighting. With unmeasured confounding, untrimmed estimates were biased irrespective of weighting method, and only Stürmer and Walker trimming consistently reduced bias. In settings where unmeasured confounding (e.g. frailty) may lead physicians to withhold treatment, Stürmer and Walker trimming should be considered before primary analysis. [ABSTRACT FROM AUTHOR]

New-user designs restricting to treatment initiators have become the preferred design for studying drug comparative safety and effectiveness using nonexperimental data. This design reduces confounding by indication and healthy-adherer bias at the cost of smaller study sizes and reduced external validity, particularly when assessing a newly approved treatment compared with standard treatment. The prevalent new-user design includes adopters of a new treatment who switched from or previously used standard treatment (i.e. the comparator), expanding study sample size and potentially broadening the study population for inference. Previous work has suggested the use of time-conditional propensity-score matching to mitigate prevalent user bias. In this study, we describe 3 "types" of initiators of a treatment: new users, direct switchers, and delayed switchers. Using these initiator types, we articulate the causal questions answered by the prevalent new-user design and compare them with those answered by the new-user design. We then show, using simulation, how conditioning on time since initiating the comparator (rather than full treatment history) can still result in a biased estimate of the treatment effect. When implemented properly, the prevalent new-user design estimates new and important causal effects distinct from the new-user design. [ABSTRACT FROM AUTHOR]

Background: Due to clinical and methodological diversity, clinical studies included in meta-analyses often differ in ways that lead to differences in treatment effects across studies. Meta-regression analysis is generally recommended to explore associations between study-level characteristics and treatment effect, however, three key pitfalls of meta-regression may lead to invalid conclusions. Our aims were to determine the frequency of these three pitfalls of meta-regression analyses, examine characteristics associated with the occurrence of these pitfalls, and explore changes between 2002 and 2012.Methods: A meta-epidemiological study of studies including aggregate data meta-regression analysis in the years 2002 and 2012. We assessed the prevalence of meta-regression analyses with at least 1 of 3 pitfalls: ecological fallacy, overfitting, and inappropriate methods to regress treatment effects against the risk of the analysed outcome. We used logistic regression to investigate study characteristics associated with pitfalls and examined differences between 2002 and 2012.Results: Our search yielded 580 studies with meta-analyses, of which 81 included meta-regression analyses with aggregated data. 57 meta-regression analyses were found to contain at least one pitfall (70%): 53 were susceptible to ecological fallacy (65%), 14 had a risk of overfitting (17%), and 5 inappropriately regressed treatment effects against the risk of the analysed outcome (6%). We found no difference in the prevalence of meta-regression analyses with methodological pitfalls between 2002 and 2012, nor any study-level characteristic that was clearly associated with the occurrence of any of the pitfalls.Conclusion: The majority of meta-regression analyses based on aggregate data contain methodological pitfalls that may result in misleading findings. [ABSTRACT FROM AUTHOR]

Epidemiologic methods have advanced tremendously in the last several decades. As important as they are, even the most sophisticated approaches are unable to provide meaningful answers when the user lacks a clear study question. Yet, instructors have more and more resources on how to conduct studies and analyze data but few resources on how to ask clearly defined study questions that will guide those methods. Training programs have limited time for coursework, and if novel statistical estimation methods become the focus of instruction, programs that go this route may end up underemphasizing the process of asking good study questions, designing robust studies, considering potential biases in the collected data, and appropriately interpreting the results of the analysis. Given the demands for space in curricula, now is an appropriate time to reevaluate what we teach epidemiology doctoral students. We advocate that programs place a renewed focus on asking good study questions and following a comprehensive approach to study design and data analysis in which questions guide the choice of appropriate methods, helping us avoid methods for methods' sake and highlighting when application of a new method can provide the opportunity to answer questions that were intractable with traditional approaches. [ABSTRACT FROM AUTHOR]

A review of epidemiological papers conducted in 2009 concluded that several studies employed variable selection methods susceptible to introduce bias and yield inadequate inferences. Many new confounder selection methods have been developed since then. The goal of the study was to provide an updated descriptive portrait of which variable selection methods are used by epidemiologists for analyzing observational data. Studies published in four major epidemiological journals in 2015 were reviewed. Only articles concerned with a predictive or explicative objective and reporting on the analysis of individual data were included. Method(s) employed for selecting variables were extracted from retained articles. A total of 975 articles were retrieved and 299 met eligibility criteria, 292 of which pursued an explicative objective. Among those, 146 studies (50%) reported using prior knowledge or causal graphs for selecting variables, 34 (12%) used change in effect estimate methods, 26 (9%) used stepwise approaches, 16 (5%) employed univariate analyses, 5 (2%) used various other methods and 107 (37%) did not provide sufficient details to allow classification (more than one method could be employed in a single article). Despite being less frequent than in the previous review, stepwise and univariable analyses, which are susceptible to introduce bias and produce inadequate inferences, were still prevalent. Moreover, 37% studies did not provide sufficient details to assess how variables were selected. We thus believe there is still room for improvement in variable selection methods used by epidemiologists and in their reporting. [ABSTRACT FROM AUTHOR]

Introduction and Aims: There is a paucity of data on volume of alcohol use from Muslim majority countries. We aimed to present estimation methods for alcohol consumption with the use of survey data for these societies and provide an estimation for age 15+ per capita consumption of pure alcohol for Iran.Design and Methods: The Iranian Mental Health Survey was a nationally representative household survey on individuals aged 15-64 years, with a multistage, cluster sampling design. We used the 'Last Week' method and 'Quantity-Frequency' methods for gathering data on alcohol consumption and combined these to provide more complete estimates.Results: The response rate was 85.7%. From the total of 7840 respondents, 5.7% and 1% reported past 12 months and past week alcohol use, respectively. The highest estimation for age 15+ per capita consumption of pure alcohol was yielded by the 'combination method' (0.108 L ethanol/person/year) followed by the Quantity-Frequency method (0.079 L). The 'Last Week' method provided the lowest estimate (0.059 L).Discussion and Conclusions: Unlike in surveys of non-Muslim countries, frequency of drinking from recent recall (last week) was much lower than from recall of usual drinking in the last year. We conclude that 0.108 L (SE = 0.03) is the best survey-based estimate of age 15+ per capita consumption, which translates to about 5 750 000 L of national consumption per year in Iran. However, this method is still likely to under-estimate per capita consumption due to evidence of under-reporting in the survey. [ABSTRACT FROM AUTHOR]

Background: National estimates of the sizes of key populations, including female sex workers, men who have sex with men, and transgender women are critical to inform national and international responses to the HIV pandemic. However, epidemiologic studies typically provide size estimates for only limited high priority geographic areas. This article illustrates a two-stage approach to obtain a national key population size estimate in the Dominican Republic using available estimates and publicly available contextual information.Methods: Available estimates of key population size in priority areas were augmented with targeted additional data collection in other areas. To combine information from data collected at each stage, we used statistical methods for handling missing data, including inverse probability weights, multiple imputation, and augmented inverse probability weights.Results: Using the augmented inverse probability weighting approach, which provides some protection against parametric model misspecification, we estimated that 3.7% (95% CI = 2.9, 4.7) of the total population of women in the Dominican Republic between the ages of 15 and 49 years were engaged in sex work, 1.2% (95% CI = 1.1, 1.3) of men aged 15-49 had sex with other men, and 0.19% (95% CI = 0.17, 0.21) of people assigned the male sex at birth were transgender.Conclusions: Viewing the size estimation of key populations as a missing data problem provides a framework for articulating and evaluating the assumptions necessary to obtain a national size estimate. In addition, this paradigm allows use of methods for missing data familiar to epidemiologists. [ABSTRACT FROM AUTHOR]

Longitudinal cohort studies can provide important evidence about preventable causes of disease, but the success relies heavily on the commitment of their participants, both at recruitment and during follow up. Initial participation rates have decreased in recent decades as have willingness to participate in subsequent follow ups. It is important to examine how such selection affects the validity of the results. In this article, we describe the conceptual framework for selection bias due to nonparticipation and loss to follow up in cohort studies, using both a traditional epidemiological approach and directed acyclic graphs. Methods to quantify selection bias are introduced together with analytical strategies to adjust for the bias including controlling for covariates associated with selection, inverse probability weighting and bias analysis. We use several studies conducted in the Danish National Birth Cohort as examples of how to quantify selection bias and also understand the underlying selection mechanisms. Although women who chose to participate in this cohort were typically of higher social status, healthier and with less disease than all those eligible for study, differential selection was modest and the influence of selection bias on several selected exposure-outcome associations was limited. These findings are reassuring and support enrolling a subset of motivated participants who would engage in long-term follow up rather than prioritize representativeness. Some of the presented methods are applicable even with limited data on nonparticipants and those lost to follow up, and can also be applied to other study designs such as case-control studies and surveys. [ABSTRACT FROM AUTHOR]

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

Confounding biases study results when the effect of the exposure on the outcome mixes with the effects of other risk and protective factors for the outcome that are present differentially by exposure status. However, not all differences between the exposed and unexposed group cause confounding. Thus, sources of confounding must be identified before they can be addressed. Confounding is absent in an ideal study where all of the population of interest is exposed in one universe and is unexposed in a parallel universe. In an actual study, an observed unexposed population represents the unobserved parallel universe. Thinking about differences between this substitute population and the unexposed parallel universe helps identify sources of confounding. These differences can then be represented in a diagram that shows how risk and protective factors for the outcome are related to the exposure. Sources of confounding identified in the diagram should be addressed analytically and through study design. However, treating all factors that differ by exposure status as confounders without considering the structure of their relation to the exposure can introduce bias. For example, conditions affected by the exposure are not confounders. There are also special types of confounding, such as time-varying confounding and unfixable confounding. It is important to evaluate carefully whether factors of interest contribute to confounding because bias can be introduced both by ignoring potential confounders and by adjusting for factors that are not confounders. The resulting bias can result in misleading conclusions about the effect of the exposure of interest on the outcome. [ABSTRACT FROM AUTHOR]

In the last few years, stakeholders in the scientific community have raised alarms about a perceived lack of reproducibility of scientific results. In reaction, guidelines for journals have been promulgated and grant applicants have been asked to address the rigor and reproducibility of their proposed projects. Neither solution addresses a primary culprit, which is the culture of null hypothesis significance testing that dominates statistical analysis and inference. In an innovative research enterprise, selection of results for further evaluation based on null hypothesis significance testing is doomed to yield a low proportion of reproducible results and a high proportion of effects that are initially overestimated. In addition, the culture of null hypothesis significance testing discourages quantitative adjustments to account for systematic errors and quantitative incorporation of prior information. These strategies would otherwise improve reproducibility and have not been previously proposed in the widely cited literature on this topic. Without discarding the culture of null hypothesis significance testing and implementing these alternative methods for statistical analysis and inference, all other strategies for improving reproducibility will yield marginal gains at best. [ABSTRACT FROM AUTHOR]

A propensity score (PS) model's ability to control confounding can be assessed by evaluating covariate balance across exposure groups after PS adjustment. The optimal strategy for evaluating a disease risk score (DRS) model's ability to control confounding is less clear. DRS models cannot be evaluated through balance checks within the full population, and they are usually assessed through prediction diagnostics and goodness-of-fit tests. A proposed alternative is the "dry-run" analysis, which divides the unexposed population into "pseudo-exposed" and "pseudo-unexposed" groups so that differences on observed covariates resemble differences between the actual exposed and unexposed populations. With no exposure effect separating the pseudo-exposed and pseudounexposed groups, a DRS model is evaluated by its ability to retrieve an unconfounded null estimate after adjustment in this pseudo-population. We used simulations and an empirical example to compare traditional DRS performance metrics with the dry-run validation. In simulations, the dry run often improved assessment of confounding control, compared with the C statistic and goodness-of-fit tests. In the empirical example, PS and DRS matching gave similar results and showed good performance in terms of covariate balance (PS matching) and controlling confounding in the dry-run analysis (DRS matching). The dry-run analysis may prove useful in evaluating confounding control through DRS models. [ABSTRACT FROM AUTHOR]

Influenza viruses undergo frequent antigenic changes. As a result, the viruses circulating change within and between seasons, and the composition of the influenza vaccine is updated annually. Thus, estimation of the vaccine's effectiveness is not constant across seasons. In order to provide annual estimates of the influenza vaccine's effectiveness, health departments have increasingly adopted the "test-negative design," using enhanced data from routine surveillance systems. In this design, patients presenting to participating general practitioners with influenza-like illness are swabbed for laboratory testing; those testing positive for influenza virus are defined as cases, and those testing negative form the comparison group. Data on patients' vaccination histories and confounder profiles are also collected. Vaccine effectiveness is estimated from the odds ratio comparing the odds of testing positive for influenza among vaccinated patients and unvaccinated patients, adjusting for confounders. The test-negative design is purported to reduce bias associated with confounding by health-care-seeking behavior and misclassification of cases. In this paper, we use directed acyclic graphs to characterize potential biases in studies of influenza vaccine effectiveness using the test-negative design. We show how studies using this design can avoid or minimize bias and where bias may be introduced with particular study design variations. [ABSTRACT FROM AUTHOR]

In time-to-event analyses of observational studies of drug effectiveness, incorrect handling of the period between cohort entry and first treatment exposure during follow-up may result in immortal time bias. This bias can be eliminated by acknowledging a change in treatment exposure status with time-dependent analyses, such as fitting a time-dependent Cox model. The prescription time-distribution matching (PTDM) method has been proposed as a simpler approach for controlling immortal time bias. Using simulation studies and theoretical quantification of bias, we compared the performance of the PTDM approach with that of the time-dependent Cox model in the presence of immortal time. Both assessments revealed that the PTDM approach did not adequately address immortal time bias. Based on our simulation results, another recently proposed observational data analysis technique, the sequential Cox approach, was found to be more useful than the PTDM approach (Cox: bias = -0.002, mean squared error = 0.025; PTDM: bias = -1.411, mean squared error = 2.011). We applied these approaches to investigate the association of β-interferon treatment with delaying disability progression in a multiple sclerosis cohort in British Columbia, Canada (Long-Term Benefits and Adverse Effects of Beta-Interferon for Multiple Sclerosis (BeAMS) Study, 1995-2008). [ABSTRACT FROM AUTHOR]

Directed acyclic graphs (DAGs) are nonparametric graphical tools used to depict causal relations in the epidemiologic assessment of exposure-outcome associations. Although their use in dental research was first advocated in 2002, DAGs have yet to be widely adopted in this field. DAGs help identify threats to causal inference such as confounders, bias due to subject selection, and inappropriate handling of missing data. DAGs can also inform the data analysis strategy based on relations among variables depicted on it. This article uses the example of a study of temporomandibular disorders (TMDs), investigating causal effects of facial injury on subsequent risk of TMD. We illustrate how DAGs can be used to identify 1) potential confounders, 2) mediators and the consequences of attempt to estimate direct causal effects, 3) colliders and the consequences of conditioning on colliders, and 4) variables that are simultaneously mediators and confounders and the consequences of adjustment for such variables. For example, one DAG shows that statistical adjustment for the pressure pain threshold would necessarily bias the causal relation between facial injury and TMD. Finally, we discuss the usefulness of DAGs during study design, subject selection, and choosing variables to be measured in a study. [ABSTRACT FROM AUTHOR]

Previous studies have compared calipers for propensity score (PS) matching, but none have considered calipers for matching on the disease risk score (DRS). We used Medicare claims data to perform 3 cohort studies of medication initiators: a study of raloxifene versus alendronate in 1 -year nonvertebral fracture risk, a study of cyclooxygenase 2 inhibitors versus nonselective nonsteroidal antiinflammatory medications in 6-month gastrointestinal bleeding, and a study of simvastatin + ezetimibe versus simvastatin alone in 6-month cardiovascular outcomes. The study periods for each cohort were 1998 through 2005,1999 through 2002, and 2004 through 2005, respectively. In each cohort, we calculated 1) a DRS, 2) a prognostic PS which included the DRS as the independent variable in a PS model, and 3) the PS for each patient. We then nearest-neighbor matched on each score in a variable ratio and a fixed ratio within 8 calipers based on the standard deviation of the logit and the natural score scale. When variable ratio matching on the DRS, a caliper of 0.05 on the natural scale performed poorly when the outcome was rare. The prognostic PS did not appear to offer any consistent practical benefits over matching on the DRS directly. In general, logit-based calipers or calipers smaller than 0.05 on the natural scale performed well when DRS matching in all examples. [ABSTRACT FROM AUTHOR]

Little is known about the heart disease risks associated with occupational, rather than traffic-related, exposure to particulate matter with aerodynamic diameter of 2.5 um or less (PM2.5). We examined long-term exposure to PM2.5 in cohorts of aluminum smelters and fabrication workers in the United States who were followed for incident ischemic heart disease from 1998 to 2012, and we addressed 2 forms of survivor bias. Left truncation bias was addressed by restricting analyses to the subcohort hired after the start of follow up. Healthy worker survivor bias, which is characterized by time-varying confounding that is affected by prior exposure, was documented only in the smelters and required the use of marginal structural Cox models. When comparing always-exposed participants above the 10th percentile of annual exposure with those below, the hazard ratios were 1.67 (95% confidence interval (CI): 1.11, 2.52) and 3.95 (95% CI: 0.87,18.00) in the full and restricted subcohorts of smelter workers, respectively. In the fabrication stratum, hazard ratios based on conditional Cox models were 0.98 (95% CI: 0.94,1.02) and 1.17 (95% CI: 1.00, 1.37) per 1 mg/m³-year in the full and restricted subcohorts, respectively. Long-term exposure to occupational PM2.5 was associated with a higher risk of ischemic heart disease among aluminum manufacturing workers, particularly in smelters, after adjustment for survivor bias. [ABSTRACT FROM AUTHOR]

Researchers conducting observational studies need to consider 3 types of biases: selection bias, information bias, and confounding bias. A whole arsenal of statistical tools can be used to deal with information and confounding biases. However, methods for addressing selection bias and unmeasured confounding are less developed. In this paper, we propose general bounding formulas for bias, including selection bias and unmeasured confounding. This should help researchers make more prudent interpretations of their (potentially biased) results. [ABSTRACT FROM AUTHOR]

Background: assessment of baseline functional status of older patients during and after intensive care unit (ICU) admission is often hampered by challenges related to the critical illness such as cognitive dysfunction, neuropsychological morbidity and pain. To explore the reliability of assessments by carefully chosen proxies, we designed a discriminating selection of proxies and evaluated agreement between patient and proxy responses by assessing activities of daily living (ADLs) at 1 month post-ICU discharge. Methods: patients ≥60 years old admitted to the medical ICU were enrolled in a prospective parent cohort studying delirium. Proxies were carefully screened at ICU admission to choose the best available respondent. Follow-up interviews, including instruments for ADLs, were conducted 1 month after ICU discharge. We examined 179 paired patient--proxy follow-up interviews. Kappa statistics assessed inter-observer agreement, and McNemar's exact test assessed response differences. Results: patients averaged 73.3 ± 8.1 years old with 29% having evidence of cognitive impairment. Proxies were most commonly spouses (38%) or children (39%). Overall, there was substantial (x≥ 0.6) to excellent agreement (x≥ 0.8) between patients and proxies on assessment of all but one basic and one instrumental ADL. Conclusion: proxies carefully chosen at ICU admission show high levels of inter-observer agreement with older patients when assessing current functional status at 1 month post-ICU discharge. This motivates further study of proxy assessments that could be used earlier in critical illness to assess premorbid functional status. [ABSTRACT FROM AUTHOR]

The consistency of propensity score (PS) estimators relies on correct specification of the PS model. The PS is frequently estimated using main-effects logistic regression. However, the underlying model assumptions may not hold. Machine learning methods provide an alternative nonparametric approach to PS estimation. In this simulation study, we evaluated the benefit of using Super Learner (SL) for PS estimation. We created 1,000 simulated data sets (n = 500) under 4 different scenarios characterized by various degrees of deviance from the usual main-term logistic regression model for the true PS. We estimated the average treatment effect using PS matching and inverse probability of treatment weighting. The estimators' performance was evaluated in terms of PS prediction accuracy, covariate balance achieved, bias, standard error, coverage, and mean squared error. All methods exhibited adequate overall balancing properties, but in the case of model misspecification, SL performed better for highly unbalanced variables. The SL-based estimators were associated with the smallest bias in cases of severe model misspecification. Our results suggest that use of SL to estimate the PS can improve covariate balance and reduce bias in a meaningful manner in cases of serious model misspecification for treatment assignment. [ABSTRACT FROM PUBLISHER]

Evidence from genetic association studies is accumulating rapidly. Field synopses have recently arisen as an unbiased way of systematically synthesizing this evidence. We performed a systematic review and appraisal of published field synopses in genetic epidemiology and assessed their main findings and methodological characteristics. We identified 61 eligible field synopses, published between January 1, 2007, and October 31, 2013, on 52 outcomes reporting 734 significant associations at the P < 0.05 level. The median odds ratio for these associations was 1.25 (interquartile range, 1.15–1.43). Egger's test was the most common method (n = 30 synopses) of assessing publication bias. Only 12 synopses (20%) used the Venice criteria to evaluate the epidemiologic credibility of their findings (n = 449 variants). Eleven synopses (18%) were accompanied by an online database that has been regularly updated. These synopses received more citations (P = 0.01) and needed a larger research team (P = 0.02) than synopses without an online database. Overall, field synopses are becoming a valuable tool for the identification of common genetic variants, especially when researchers follow relevant methodological guidelines. Our work provides a summary of the current status of the field synopses published to date and may help interested readers efficiently identify the online resources containing the relevant genetic evidence. [ABSTRACT FROM PUBLISHER]

As highlighted in a recent editorial in the Journal (Am J Epidemiol. 2014;180(2):127–128), the research area of “-omics” includes genomics, proteomics, metabolomics, and nascent fields of scientific inquiry such as epigenomics and exposomics. These fields can be collectively referred to as “global -omics.” Increasing efforts have been made over the past 2 decades to identify and modify environmental risk factors among persons who are susceptible to disease because of their genotype and to integrate genetic information and other biological variables with information about individual-level risk factors and group-level or societal factors related to the broader residential, behavioral, or cultural context. In genome-wide association studies, only a small proportion of heritability is explained by genetic variants identified to date, which has prompted researchers in bioinformatics and biostatistics to take into account nonlinear relationships due to gene-environment or gene-gene interactions. The exposome, which is dynamic and variable, consists of all of the internal and external exposures an individual incurs over a lifetime. Both the epigenome and exposome change with age. The prenatal and perinatal periods are thought to be important times for epigenetic marking. Once the human epigenome has been fully mapped, identification of the effects of all deleterious environmental exposures according to duration of exposure and time period will be a complex undertaking, requiring collaborative epidemiologic studies. [ABSTRACT FROM PUBLISHER]

In this issue of the Journal, an expert panel offers 7 recommendations on how population studies supported by National Heart, Lung, and Blood Institute contracts might be strategically transformed (Am J Epidemiol. 2015;181(6):363–368). The Institute and its external advisors seemingly established this panel of epidemiologists and nonepidemiologists primarily to find ways to save research costs. Although the working group's recommendations offer reasonable approaches, we believe that, even in tough fiscal times, the main drivers of cardiovascular epidemiologic research must remain 1) scientific questions that are important and 2) study designs to match these. Although cardiovascular epidemiology admittedly is often redundant and needs to be more efficient, undue focus on administrative efficiency and cost savings will not necessarily guarantee cutting-edge population research. [ABSTRACT FROM PUBLISHER]

Background: Value of information is now recognized as a reference method in the decision process underpinning cost-effectiveness evaluation. The expected value of perfect information (EVPI) is the expected value from completely reducing the uncertainty surrounding the cost-effectiveness of an innovative intervention. Among sample size calculation methods used in cost-effectiveness studies, only one is coherent with this decision framework. It uses a Bayesian approach and requires data of a pre-existing cost-effectiveness study to derive a valid prior EVPI. When evaluating the cost-effectiveness of innovations, no observed prior EVPI is usually available to calculate the sample size. We here propose a sample size calculation method for cost-effectiveness studies, that follows the value of information theory, and, being frequentist, can be based on assumptions if no observed prior EVPI is available.Methods: The general principle of our method is to define the sampling distribution of the incremental net monetary benefit (ΔB), or the distribution of ΔB that would be observed in a planned cost-effectiveness study of size n. Based on this sampling distribution, the EVPI that would remain at the end of the trial (EVPIn) is estimated. The optimal sample size of the planned cost-effectiveness study is the n for which the cost of including an additional participant becomes equal or higher than the value of the information gathered through this inclusion.Results: Our method is illustrated through four examples. The first one is used to present the method in depth and describe how the sample size may vary according to the parameters' value. The three other examples are used to illustrate in different situations how the sample size may vary according to the ceiling cost-effectiveness ratio, and how it compares with a test statistic-based method. We developed an R package (EBASS) to run these calculations.Conclusions: Our sample size calculation method follows the value of information theory that is now recommended for analyzing and interpreting cost-effectiveness data, and sets the size of a study that balances its cost and the value of its information. [ABSTRACT FROM AUTHOR]