Dataset Information

Selection Bias When Estimating Average Treatment Effects Using One-sample Instrumental Variable Analysis.

ABSTRACT: Participants in epidemiologic and genetic studies are rarely true random samples of the populations they are intended to represent, and both known and unknown factors can influence participation in a study (known as selection into a study). The circumstances in which selection causes bias in an instrumental variable (IV) analysis are not widely understood by practitioners of IV analyses. We use directed acyclic graphs (DAGs) to depict assumptions about the selection mechanism (factors affecting selection) and show how DAGs can be used to determine when a two-stage least squares IV analysis is biased by different selection mechanisms. Through simulations, we show that selection can result in a biased IV estimate with substantial confidence interval (CI) undercoverage, and the level of bias can differ between instrument strengths, a linear and nonlinear exposure-instrument association, and a causal and noncausal exposure effect. We present an application from the UK Biobank study, which is known to be a selected sample of the general population. Of interest was the causal effect of staying in school at least 1 extra year on the decision to smoke. Based on 22,138 participants, the two-stage least squares exposure estimates were very different between the IV analysis ignoring selection and the IV analysis which adjusted for selection (e.g., risk differences, 1.8% [95% CI, -1.5%, 5.0%] and -4.5% [95% CI, -6.6%, -2.4%], respectively). We conclude that selection bias can have a major effect on an IV analysis, and further research is needed on how to conduct sensitivity analyses when selection depends on unmeasured data.

SUBMITTER: Hughes RA

PROVIDER: S-EPMC6525095 | biostudies-literature | 2019 May

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Selection Bias When Estimating Average Treatment Effects Using One-sample Instrumental Variable Analysis.

Hughes Rachael A RA Davies Neil M NM Davey Smith George G Tilling Kate K

Epidemiology (Cambridge, Mass.) 20190501 3

Participants in epidemiologic and genetic studies are rarely true random samples of the populations they are intended to represent, and both known and unknown factors can influence participation in a study (known as selection into a study). The circumstances in which selection causes bias in an instrumental variable (IV) analysis are not widely understood by practitioners of IV analyses. We use directed acyclic graphs (DAGs) to depict assumptions about the selection mechanism (factors affecting ...[more]

PMID: 30896457

Similar Datasets

Project description:OBJECTIVES:To assess the incremental effects of adding extra antihypertensive drugs from a new class to a patient's regimen. DESIGN:Instrumental variable analysis of data from SPRINT (Systolic Blood Pressure Intervention Trial). To account for confounding by indication-when treatments seem less effective if they are administered to sicker patients-randomization status was used as the instrumental variable. Patients' randomization status was either intensive (systolic blood pressure target <120 mm Hg) or standard (systolic blood pressure target <140 mm Hg) treatment. Results from instrumental variable models were compared with those from standard multivariable models. SETTING:Secondary data analysis of a randomized clinical trial conducted at 102 sites in 2010-15. PARTICIPANTS:9092 SPRINT participants with hypertension and increased cardiovascular risk but no history of diabetes or stroke. MAIN OUTCOMES MEASURES:Systolic blood pressure, major cardiovascular events, and serious adverse events. RESULTS:In standard multivariable models not adjusted for confounding by indication, addition of an antihypertensive drug from a new class was associated with modestly lower systolic blood pressure (-1.3 mm Hg, 95% confidence interval -1.6 to -1.0) and no change in major cardiovascular events (absolute risk of events per 1000 patient years, 0.5, 95% confidence interval -1.5 to 2.3). In instrumental variable models, the addition of an antihypertensive drug from a new class led to clinically important reductions in systolic blood pressure (-14.4 mm Hg, -15.6 to -13.3) and fewer major cardiovascular events (absolute risk -6.2, -10.9 to -1.3). Incremental reductions in systolic blood pressure remained large and similar in magnitude for patients already taking drugs from zero, one, two, or three or more drug classes. This finding was consistent across all subgroups of patients. The addition of another antihypertensive drug class was not associated with adverse events in either standard or instrumental variable models. CONCLUSIONS:After adjustment for confounding by indication, the addition of a new antihypertensive drug class led to large reductions in systolic blood pressure and major cardiovascular events among patients at high risk for cardiovascular events but without diabetes. Effects on systolic blood pressure persisted across all levels of baseline drug use and all subgroups of patients.

Dataset Information

Selection Bias When Estimating Average Treatment Effects Using One-sample Instrumental Variable Analysis.

Publications

Selection Bias When Estimating Average Treatment Effects Using One-sample Instrumental Variable Analysis.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets