The role of matching when adjusting for baseline differences in the outcome variable of comparative effectiveness studies.
Ontology highlight
ABSTRACT: Aim
Evaluate performance of analytical strategies commonly used to adjust for baseline differences in continuous outcome variables for comparative effectiveness studies.Patients & methods
Data simulations resembling a comparison of HbA1c values after initiation of antidiabetic treatments adjusting for baseline HbA1c. We evaluated change scores, analyses of covariance including linear, nonlinear with/without robust variance estimations, before and after optimal matching. We also evaluated the impact of measurement error.Results
With increasing HbA1c baseline differences between groups, bias in effect estimates and suboptimal CI coverage probabilities increased in all approaches. These issues were further compounded by measurement error. Matching on baseline HbA1c, substantially mitigated these issues.Conclusion
In comparative studies with continuous outcomes, matching on baseline values of the outcome variable improves analytical performance.
SUBMITTER: Grijalva CG
PROVIDER: S-EPMC4699664 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:BACKGROUND: Area-level variation in treatment and outcomes may be a potential source of confounding bias in observational comparative effectiveness studies. This paper demonstrates how to use exploratory spatial data analysis (ESDA) and spatial statistical methods to investigate and control for these potential biases. The case presented compares the effectiveness of two antipsychotic treatment strategies: oral second-generation antipsychotics (SGAs) vs. long-acting paliperiodone palmitate (PP). METHODS: A new-start cohort study was conducted analyzing patient-level administrative claims data (8/1/2008-4/30/2011) from Missouri Medicaid. ESDA techniques were used to examine spatial patterns of antipsychotic prescriptions and outcomes (hospitalization and emergency department (ED) visits). Likelihood of mental health-related outcomes were compared between patients starting PP (N?=?295) and oral SGAs (N?=?8,626) using multilevel logistic regression models adjusting for patient composition (demographic and clinical factors) and geographic region. RESULTS: ESDA indicated significant spatial variation in antipsychotic prescription patterns and moderate variation in hospitalization and ED visits thereby indicating possible confounding by geography. In the multilevel models for this antipsychotic case example, patient composition represented a stronger source of confounding than geographic context. CONCLUSION: Because geographic variation in health care delivery is ubiquitous, it could be a comparative effectiveness research (CER) best practice to test for possible geographic confounding in observational data. Though the magnitude of the area-level geography effects were small in this case, they were still statistically significant and should therefore be examined as part of this observational CER study. More research is needed to better estimate the range of confounding due to geography across different types of observational comparative effectiveness studies and healthcare utilization outcomes.
| S-EPMC4161848 | biostudies-other
Project description:We developed a three-step matching algorithm to enhance the between-group comparability for comparative drug effect studies involving prevalent new-users of the newer study drug versus older comparator drug(s). The three-step matching scheme is to match on: (1) index date of initiating the newer study drug to align the cohort entry time between study groups, (2) medication possession ratio measures that consider prior exposure to all older comparator drugs, and (3) propensity scores estimated from potential confounders. Our approach is illustrated with a comparative cardiovascular safety study of glucagon-like peptide-1 receptor agonist (GLP-1ra) versus sulfonylurea (SU) in type 2 diabetes patients using Taiwan's National Health Insurance Research Database 2003-2015. 66% of 3195 GLP-1ra users had previously exposed to SU. The between-group comparability was well-achieved after implementing the matching algorithm (i.e., standardized mean difference < 0.2 for all baseline patient characteristics). Compared to SU, the use of GLP-1ra yielded a significantly reduced risk of the primary composite cardiovascular events (hazard ratio [95% confidence interval]: 0.71 [0.54-0.95], p = 0.022). Our matching scheme can enhance the between-group comparability in prevalent new-user cohort designs to minimize time-related bias, improve confounder adjustment, and ensure the reliability and validity of study findings.
| S-EPMC8741962 | biostudies-literature
Project description:In pharmacogenomic studies of quantitative change, any association between genetic variants and the pretreatment (baseline) measurement can bias the estimate of effect between those variants and drug response. A putative solution is to adjust for baseline. We conducted a series of genome-wide association studies (GWASs) for low-density lipoprotein cholesterol (LDL-C) response to statin therapy in 34,874 participants of the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort as a case study to investigate the impact of baseline adjustment on results generated from pharmacogenomic studies of quantitative change. Across phenotypes of statin-induced LDL-C change, baseline adjustment identified variants from six loci meeting genome-wide significance (SORT/CELSR2/PSRC1, LPA, SLCO1B1, APOE, APOB, and SMARCA4/LDLR). In contrast, baseline-unadjusted analyses yielded variants from three loci meeting the criteria for genome-wide significance (LPA, APOE, and SLCO1B1). A genome-wide heterogeneity test of baseline versus statin on-treatment LDL-C levels was performed as the definitive test for the true effect of genetic variants on statin-induced LDL-C change. These findings were generally consistent with the models not adjusting for baseline signifying that genome-wide significant hits generated only from baseline-adjusted analyses (SORT/CELSR2/PSRC1, APOB, SMARCA4/LDLR) were likely biased. We then comprehensively reviewed published GWASs of drug-induced quantitative change and discovered that more than half (59%) inappropriately adjusted for baseline. Altogether, we demonstrate that (1) baseline adjustment introduces bias in pharmacogenomic studies of quantitative change and (2) this erroneous methodology is highly prevalent. We conclude that it is critical to avoid this common statistical approach in future pharmacogenomic studies of quantitative change.
| S-EPMC6965183 | biostudies-literature
Project description:BackgroundComparative-effectiveness studies using real-world data (RWD) can be susceptible to surveillance bias. In solid tumor oncology studies, analyses of endpoints such as progression-free survival (PFS) are based on progression events detected by imaging assessments. This study aimed to evaluate the potential bias introduced by differential imaging assessment frequency when using electronic health record (EHR)-derived data to investigate the comparative effectiveness of cancer therapies.MethodsUsing a nationwide de-identified EHR-derived database, we first analyzed imaging assessment frequency patterns in patients diagnosed with advanced non-small cell lung cancer (aNSCLC). We used those RWD inputs to develop a discrete event simulation model of two treatments where disease progression was the outcome and PFS was the endpoint. Using this model, we induced bias with differential imaging assessment timing and quantified its effect on observed versus true treatment effectiveness. We assessed percent bias in the estimated hazard ratio (HR).ResultsThe frequency of assessments differed by cancer treatment types. In simulated comparative-effectiveness studies, PFS HRs estimated using real-world imaging assessment frequencies differed from the true HR by less than 10% in all scenarios (range: 0.4% to -9.6%). The greatest risk of biased effect estimates was found comparing treatments with widely different imaging frequencies, most exaggerated in disease settings where time to progression is very short.ConclusionsThis study provided evidence that the frequency of imaging assessments to detect disease progression can differ by treatment type in real-world patients with cancer and may induce some bias in comparative-effectiveness studies in some situations.
| S-EPMC9290806 | biostudies-literature
Project description:In studies that compare several diagnostic or treatment groups, subjects may not only be measured on a certain set of feature variables, but also be matched on a number of demographic characteristics and measured on additional covariates. Linear discriminant analysis (LDA) is sometimes used to identify which feature variables best discriminate among groups, while accounting for the dependencies among the feature variables. We present a new approach to LDA for multivariate normal data that accounts for the subject matching used in a particular study design, as well as covariates not used in the matching. Applications are given for post-mortem tissue data with the aim of comparing neurobiological characteristics of subjects with schizophrenia with those of normal controls, and for a post-mortem tissue primate study comparing brain biomarker measurements across three treatment groups. We also investigate the performance of our approach using a simulation study.
| S-EPMC3770000 | biostudies-literature
Project description:Case-cohort designs are widely used in large cohort studies to reduce the cost associated with covariate measurement. In many such studies the number of covariates is very large, so an efficient variable selection method is necessary. In this paper, we study the properties of a variable selection procedure using the smoothly clipped absolute deviation penalty in a case-cohort design with a diverging number of parameters. We establish the consistency and asymptotic normality of the maximum penalized pseudo-partial-likelihood estimator, and show that the proposed variable selection method is consistent and has an asymptotic oracle property. Simulation studies compare the finite-sample performance of the procedure with tuning parameter selection methods based on the Akaike information criterion and the Bayesian information criterion. We make recommendations for use of the proposed procedures in case-cohort studies, and apply them to the Busselton Health Study.
| S-EPMC5436496 | biostudies-literature
Project description:BackgroundDetermination of comparative effectiveness in a randomized controlled trial requires consideration of an intervention's comparative uptake (or acceptance) among randomized participants and the intervention's comparative efficacy among participants who use their assigned intervention. If acceptance differs across interventions, then simple randomization of participants can result in post-randomization losses that introduce bias and limit statistical power.MethodsWe develop a novel preference-adaptive randomization procedure in which the allocation probabilities are updated based on the inverse of the relative acceptance rates among randomized participants in each arm. In simulation studies, we determine the optimal frequency with which to update the allocation probabilities based on the number of participants randomized. We illustrate the development and application of preference-adaptive randomization using a randomized controlled trial comparing the effectiveness of different financial incentive structures on prolonged smoking cessation.ResultsSimulation studies indicated that preference-adaptive randomization performed best with frequent updating, accommodated differences in acceptance across arms, and performed well even if the initial values for the allocation probabilities were not equal to their true values. Updating the allocation probabilities after randomizing each participant minimized imbalances in the number of accepting participants across arms over time. In the smoking cessation trial, unexpectedly large differences in acceptance among arms required us to limit the allocation of participants to less acceptable interventions. Nonetheless, the procedure achieved equal numbers of accepting participants in the more acceptable arms, and balanced the characteristics of participants across assigned interventions.ConclusionsPreference-adaptive randomization, coupled with analysis methods based on instrumental variables, can enhance the validity and generalizability of comparative effectiveness studies. In particular, preference-adaptive randomization augments statistical power by maintaining balanced sample sizes in efficacy analyses, while retaining the ability of randomization to balance covariates across arms in effectiveness analyses.Trial registrationClinicalTrials.gov, NCT01526265; 31 January 2012.
| S-EPMC4387665 | biostudies-literature
Project description:Estimation of causal effects of time-varying exposures using longitudinal data is a common problem in epidemiology. When there are time-varying confounders, which may include past outcomes, affected by prior exposure, standard regression methods can lead to bias. Methods such as inverse probability weighted estimation of marginal structural models have been developed to address this problem. However, in this paper we show how standard regression methods can be used, even in the presence of time-dependent confounding, to estimate the total effect of an exposure on a subsequent outcome by controlling appropriately for prior exposures, outcomes, and time-varying covariates. We refer to the resulting estimation approach as sequential conditional mean models (SCMMs), which can be fitted using generalized estimating equations. We outline this approach and describe how including propensity score adjustment is advantageous. We compare the causal effects being estimated using SCMMs and marginal structural models, and we compare the two approaches using simulations. SCMMs enable more precise inferences, with greater robustness against model misspecification via propensity score adjustment, and easily accommodate continuous exposures and interactions. A new test for direct effects of past exposures on a subsequent outcome is described.
| S-EPMC5928464 | biostudies-literature
Project description:Propensity score matching is a common tool for adjusting for observed confounding in observational studies, but is known to have limitations in the presence of unmeasured confounding. In many settings, researchers are confronted with spatially-indexed data where the relative locations of the observational units may serve as a useful proxy for unmeasured confounding that varies according to a spatial pattern. We develop a new method, termed distance adjusted propensity score matching (DAPSm) that incorporates information on units' spatial proximity into a propensity score matching procedure. We show that DAPSm can adjust for both observed and some forms of unobserved confounding and evaluate its performance relative to several other reasonable alternatives for incorporating spatial information into propensity score adjustment. The method is motivated by and applied to a comparative effectiveness investigation of power plant emission reduction technologies designed to reduce population exposure to ambient ozone pollution. Ultimately, DAPSm provides a framework for augmenting a "standard" propensity score analysis with information on spatial proximity and provides a transparent and principled way to assess the relative trade-offs of prioritizing observed confounding adjustment versus spatial proximity adjustment.
| S-EPMC6409420 | biostudies-literature
Project description:Nursing home residents are of particular interest for comparative effectiveness research given their susceptibility to adverse treatment effects and systematic exclusion from trials. However, the risk of residual confounding because of unmeasured markers of declining health using conventional analytic methods is high. We evaluated the validity of instrumental variable (IV) methods based on nursing home prescribing preference to mitigate such confounding, using psychotropic medications to manage behavioral problems in dementia as a case study.A cohort using linked data from Medicaid, Medicare, Minimum Data Set, and Online Survey, Certification and Reporting for 2001-2004 was established. Dual-eligible patients ?65?years who initiated psychotropic medication use after admission were selected. Nursing home prescribing preference was characterized using mixed-effects logistic regression models. The plausibility of IV assumptions was explored, and the association between psychotropic medication class and 180-day mortality was estimated.High-prescribing and low-prescribing nursing homes differed by a factor of 2. Each preference-based IV measure described a substantial proportion of variation in psychotropic medication choice (?(IV???treatment): 0.22-0.36). Measured patient characteristics were well balanced across patient groups based on instrument status (52% average reduction in Mahalanobis distance). There was no evidence that instrument status was associated with markers of nursing home quality of care.Findings indicate that IV analyses using nursing home prescribing preference may be a useful approach in comparative effectiveness studies, and should extend naturally to analyses including untreated comparison groups, which are of great scientific interest but subject to even stronger confounding.
| S-EPMC4116440 | biostudies-literature