Comparing methods to address bias in observational data

Statin use and cardiovascular events in a US cohort

Paulina Kaiser, Alice M. Arnold, David Benkeser, Adina Zeki Al Hazzouri, Calvin H. Hirsch, Bruce M. Psaty, Michelle C. Odden

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Background: The theoretical conditions under which causal estimates can be derived from observational data are challenging to achieve in the real world. Applied examples can help elucidate the practical limitations of methods to estimate randomized- controlled trial effects from observational data. Methods: We used six methods with varying design and analytic features to compare the 5-year risk of incident myocardial infarction among statin users and non-users, and used non-cardiovascular mortality as a negative control outcome. Design features included restriction to a statin-eligible population and new users only; analytic features included multivariable adjustment and propensity score matching. Results: We used data from 5294 participants in the Cardiovascular Health Study from 1989 to 2004. For non-cardiovascular mortality, most methods produced protective estimates with confidence intervals that crossed the null. The hazard ratio (HR) was 0.92, 95% confidence interval: 0.58, 1.46 using propensity score matching among eligible new users. For myocardial infarction, all estimates were strongly protective; the propensity score-matched analysis among eligible new users resulted in a HR of 0.55 (0.29, 1.05)-a much stronger association than observed in randomized controlled trials. Conclusions: In designs that compare active treatment with non-treated participants to evaluate effectiveness, methods to address bias in observational data may be limited in real-world settings by residual bias.

Original languageEnglish (US)
Article numberdyx217
Pages (from-to)246-254
Number of pages9
JournalInternational Journal of Epidemiology
Volume47
Issue number1
DOIs
StatePublished - Feb 1 2018

Fingerprint

Hydroxymethylglutaryl-CoA Reductase Inhibitors
Propensity Score
Randomized Controlled Trials
Myocardial Infarction
Confidence Intervals
Mortality
Health
Population

Keywords

  • Bias
  • Observational studies
  • Statins

ASJC Scopus subject areas

  • Epidemiology

Cite this

Comparing methods to address bias in observational data : Statin use and cardiovascular events in a US cohort. / Kaiser, Paulina; Arnold, Alice M.; Benkeser, David; Zeki Al Hazzouri, Adina; Hirsch, Calvin H.; Psaty, Bruce M.; Odden, Michelle C.

In: International Journal of Epidemiology, Vol. 47, No. 1, dyx217, 01.02.2018, p. 246-254.

Research output: Contribution to journalArticle

Kaiser, Paulina ; Arnold, Alice M. ; Benkeser, David ; Zeki Al Hazzouri, Adina ; Hirsch, Calvin H. ; Psaty, Bruce M. ; Odden, Michelle C. / Comparing methods to address bias in observational data : Statin use and cardiovascular events in a US cohort. In: International Journal of Epidemiology. 2018 ; Vol. 47, No. 1. pp. 246-254.
@article{d2e7f20d9c9a4dedb225bff2a525d0b6,
title = "Comparing methods to address bias in observational data: Statin use and cardiovascular events in a US cohort",
abstract = "Background: The theoretical conditions under which causal estimates can be derived from observational data are challenging to achieve in the real world. Applied examples can help elucidate the practical limitations of methods to estimate randomized- controlled trial effects from observational data. Methods: We used six methods with varying design and analytic features to compare the 5-year risk of incident myocardial infarction among statin users and non-users, and used non-cardiovascular mortality as a negative control outcome. Design features included restriction to a statin-eligible population and new users only; analytic features included multivariable adjustment and propensity score matching. Results: We used data from 5294 participants in the Cardiovascular Health Study from 1989 to 2004. For non-cardiovascular mortality, most methods produced protective estimates with confidence intervals that crossed the null. The hazard ratio (HR) was 0.92, 95{\%} confidence interval: 0.58, 1.46 using propensity score matching among eligible new users. For myocardial infarction, all estimates were strongly protective; the propensity score-matched analysis among eligible new users resulted in a HR of 0.55 (0.29, 1.05)-a much stronger association than observed in randomized controlled trials. Conclusions: In designs that compare active treatment with non-treated participants to evaluate effectiveness, methods to address bias in observational data may be limited in real-world settings by residual bias.",
keywords = "Bias, Observational studies, Statins",
author = "Paulina Kaiser and Arnold, {Alice M.} and David Benkeser and {Zeki Al Hazzouri}, Adina and Hirsch, {Calvin H.} and Psaty, {Bruce M.} and Odden, {Michelle C.}",
year = "2018",
month = "2",
day = "1",
doi = "10.1093/ije/dyx179",
language = "English (US)",
volume = "47",
pages = "246--254",
journal = "International Journal of Epidemiology",
issn = "0300-5771",
publisher = "Oxford University Press",
number = "1",

}

TY - JOUR

T1 - Comparing methods to address bias in observational data

T2 - Statin use and cardiovascular events in a US cohort

AU - Kaiser, Paulina

AU - Arnold, Alice M.

AU - Benkeser, David

AU - Zeki Al Hazzouri, Adina

AU - Hirsch, Calvin H.

AU - Psaty, Bruce M.

AU - Odden, Michelle C.

PY - 2018/2/1

Y1 - 2018/2/1

N2 - Background: The theoretical conditions under which causal estimates can be derived from observational data are challenging to achieve in the real world. Applied examples can help elucidate the practical limitations of methods to estimate randomized- controlled trial effects from observational data. Methods: We used six methods with varying design and analytic features to compare the 5-year risk of incident myocardial infarction among statin users and non-users, and used non-cardiovascular mortality as a negative control outcome. Design features included restriction to a statin-eligible population and new users only; analytic features included multivariable adjustment and propensity score matching. Results: We used data from 5294 participants in the Cardiovascular Health Study from 1989 to 2004. For non-cardiovascular mortality, most methods produced protective estimates with confidence intervals that crossed the null. The hazard ratio (HR) was 0.92, 95% confidence interval: 0.58, 1.46 using propensity score matching among eligible new users. For myocardial infarction, all estimates were strongly protective; the propensity score-matched analysis among eligible new users resulted in a HR of 0.55 (0.29, 1.05)-a much stronger association than observed in randomized controlled trials. Conclusions: In designs that compare active treatment with non-treated participants to evaluate effectiveness, methods to address bias in observational data may be limited in real-world settings by residual bias.

AB - Background: The theoretical conditions under which causal estimates can be derived from observational data are challenging to achieve in the real world. Applied examples can help elucidate the practical limitations of methods to estimate randomized- controlled trial effects from observational data. Methods: We used six methods with varying design and analytic features to compare the 5-year risk of incident myocardial infarction among statin users and non-users, and used non-cardiovascular mortality as a negative control outcome. Design features included restriction to a statin-eligible population and new users only; analytic features included multivariable adjustment and propensity score matching. Results: We used data from 5294 participants in the Cardiovascular Health Study from 1989 to 2004. For non-cardiovascular mortality, most methods produced protective estimates with confidence intervals that crossed the null. The hazard ratio (HR) was 0.92, 95% confidence interval: 0.58, 1.46 using propensity score matching among eligible new users. For myocardial infarction, all estimates were strongly protective; the propensity score-matched analysis among eligible new users resulted in a HR of 0.55 (0.29, 1.05)-a much stronger association than observed in randomized controlled trials. Conclusions: In designs that compare active treatment with non-treated participants to evaluate effectiveness, methods to address bias in observational data may be limited in real-world settings by residual bias.

KW - Bias

KW - Observational studies

KW - Statins

UR - http://www.scopus.com/inward/record.url?scp=85039712687&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85039712687&partnerID=8YFLogxK

U2 - 10.1093/ije/dyx179

DO - 10.1093/ije/dyx179

M3 - Article

VL - 47

SP - 246

EP - 254

JO - International Journal of Epidemiology

JF - International Journal of Epidemiology

SN - 0300-5771

IS - 1

M1 - dyx217

ER -