Biased Estimates in Mendelian Randomization Studies Conducted in Unrepresentative Samples

Biased Estimates in Mendelian Randomization Studies Conducted in Unrepresentative Samples Letters 3 4 We agree that a CAC score of 0 in symptomatic patients may Lyall et al use mendelian randomization, an increasingly not represent low risk. On the other hand, as we showed in popular method to support stronger causal inferences in obser- healthy asymptomatic older people where risk assessment and vational data, but this does not protect against the potential bias drugprescriptionisthemainconsideration,aCACscoreof0most we describe. In the Avon Longitudinal Study of Parents and Chil- likely represents a much lower-risk individual. dren (ALSPAC), a large prospective cohort study in the United Kingdom, both body mass index and smoking predict atten- Philip Greenland, MD dance at subsequent assessment clinics. Genetic antecedents Yuichiro Yano, MD, PhD of factors, such as body mass index and smoking that influence Donald M. Lloyd-Jones, MD, ScM participation, will be subject to a similar bias, which we see in the ALSPAC. Author Affiliations: Department of Preventive Medicine, Northwestern The extent to which bias owing to selection will occur will University Feinberg School of Medicine, Chicago, Illinois (Greenland, Lloyd- depend on the particular association being explored and the op- Jones); Department of Preventive Medicine, University of Mississippi Medical erating selection mechanisms. We suggest that researchers con- Center, Jackson (Yano). sider the potential for selection bias to be affecting their analy- Corresponding Author: Philip Greenland, MD, Department of Preventive Medicine, Northwestern University Feinberg School of Medicine, 680 N Lake ses and carry out sensitivity analyses to assess robustness of their Shore Dr, Ste 1400, Chicago, IL 60611 (p-greenland@northwestern.edu). conclusions to selection bias. Published Online: November 29, 2017. doi:10.1001/jamacardio.2017.4350 Conflict of Interest Disclosures: All authors have completed and submitted the Marcus Munafò, PhD ICMJE Form for Disclosure of Potential Conflicts of Interest and none were George Davey Smith, DSc reported. 1. Yano Y, O’Donnell CJ, Kuller L, et al. Association of coronary artery calcium Author Affiliations: MRC Integrative Epidemiology Unit, University of Bristol, score vs age with cardiovascular risk in older adults: an analysis of pooled Bristol, United Kingdom (Munafò, Smith); UK Centre for Tobacco and Alcohol population-based studies. JAMA Cardiol. 2017;2(9):986-994. Studies, School of Experimental Psychology, University of Bristol, Bristol, United 2. Schenker MP, Dorbala S, Hong EC, et al. Interrelation of coronary Kingdom (Munafò); School of Social and Community Medicine, University of calcification, myocardial ischemia, and outcomes in patients with intermediate Bristol, Bristol, United Kingdom (Smith). likelihood of coronary artery disease: a combined positron emission Corresponding Author: Marcus Munafò, PhD, UK Centre for Tobacco and tomography/computed tomography study. Circulation. 2008;117(13):1693-1700. Alcohol Studies, School of Experimental Psychology, University of Bristol, 12a 3. Greenland P, Bonow RO. How low-risk is a coronary calcium score of zero? Priory Rd, Bristol BS8 1TU, United Kingdom (marcus.munafo@bristol.ac.uk). the importance of conditional probability. Circulation. 2008;117(13):1627-1629. Published Online: December 13, 2017. doi:10.1001/jamacardio.2017.4279 Conflict of Interest Disclosures: Both authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were Biased Estimates in Mendelian Randomization reported. Studies Conducted in Unrepresentative Samples Editorial Note: This letter was shown to the corresponding author of the original article, who declined to reply on behalf of the authors. To the Editor Large cohort studies have transformed our ability 1. Fry A, Littlejohns TJ, Sudlow C, et al. Comparison of sociodemographic and to evaluate complex exposure-outcome relationships. However, health-related characteristics of UK biobank participants with the general these studies are typically not representative of the source popu- population [published online June 21, 2017]. Am J Epidemiol. lation from which they are sampled because of selection at re- 2. Munafò MR, Tilling K, Taylor AE, Evans DM, Davey Smith G. Collider scope: cruitment, attrition over time (in prospective studies), or both. when selection bias can substantially influence observed associations Forexample,inUKBiobank,theresponseratewasapproximately [published online June 7, 2017]. Int J Epidemiol. doi:10.1101/079707 5%, and the resulting sample was healthier and more highly edu- 3. Lyall DM, Celis-Morales C, Ward J, et al. Association of body mass index with cardiometabolic disease in the UK biobank: a Mendelian randomization study. cated than the general population of the United Kingdom. JAMA Cardiol. 2017;2(8):882-889. It is often assumed that while this lack of representativeness 4. Davey Smith G, Ebrahim S. ‘Mendelian randomization’: can genetic is problematic for estimating prevalence, it will make little or no epidemiology contribute to understanding environmental determinants of difference to association analyses. In our view, this is overly op- disease? Int J Epidemiol. 2003;32(1):1-22. timistic. In 2017, we examined the potential effect of selection bias on results obtained from studies with low response rates. We argue that because selection can induce collider bias (which CORRECTION occurs when 2 variables independently influence a third variable Error in Abstract: In the Original Investigation by Pokharel et al titled “Association and that variable is conditioned on), selection can lead to biased ofSerialKansasCityCardiomyopathyQuestionnaireAssessmentsWithDeathandHos- estimates of associations. pitalizationinPatientsWithHeartFailureWithPreservedandReducedEjectionFrac- tion: A Secondary Analysis of 2 Randomized Clinical Trials,” published online Novem- In this context, it is worth considering the analysis in JAMA 3 ber 1, 2017, there was an error in the Results subsection of the Abstract. The first Cardiology by Lyall et al, who reported an association of body sentenceshouldbereplacedwith2sentencesthatread“Of1767eligibleTOPCATpar- mass index with cardiometabolic disease in UK Biobank. The ex- ticipants, 882 were women (49.9%), and the mean (SD) age was 71.5 (9.7) years. Of posure and several of the outcomes (eg, coronary heart disease 2130 eligible HF-ACTION participants, 599 were women (28.1%), and the mean age was 58.6 (12.7) years.” This article was corrected online. and type 2 diabetes) might plausibly be expected to be negatively 1. Pokharel Y, Khariton Y, Tang Y, et al. Association of serial Kansas City associatedwithparticipationinUKBiobank;inwhichcase,aspu- Cardiomyopathy Questionnaire assessments with death and hospitalization in rious negative association between the 2 will be created, and any patients with heart failure with preserved and reduced ejection fraction: true association will be biased, because the use of a sample sub- a secondary analysis of 2 randomized clinical trials [published online November ject to this selection amounts to conditioning on a collider. 1, 2017]. JAMA Cardiol. doi:10.1001/jamacardio.2017.3983 jamacardiology.com (Reprinted) JAMA Cardiology February 2018 Volume 3, Number 2 181 © 2017 American Medical Association. All rights reserved. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JAMA Cardiology American Medical Association

Biased Estimates in Mendelian Randomization Studies Conducted in Unrepresentative Samples

Free
1 page

Loading next page...
1 Page
 
/lp/jama/biased-estimates-in-mendelian-randomization-studies-conducted-in-CtZAaq5CIr
Publisher
American Medical Association
Copyright
Copyright 2017 American Medical Association. All Rights Reserved.
ISSN
2380-6583
eISSN
2380-6591
D.O.I.
10.1001/jamacardio.2017.4279
Publisher site
See Article on Publisher Site

Abstract

Letters 3 4 We agree that a CAC score of 0 in symptomatic patients may Lyall et al use mendelian randomization, an increasingly not represent low risk. On the other hand, as we showed in popular method to support stronger causal inferences in obser- healthy asymptomatic older people where risk assessment and vational data, but this does not protect against the potential bias drugprescriptionisthemainconsideration,aCACscoreof0most we describe. In the Avon Longitudinal Study of Parents and Chil- likely represents a much lower-risk individual. dren (ALSPAC), a large prospective cohort study in the United Kingdom, both body mass index and smoking predict atten- Philip Greenland, MD dance at subsequent assessment clinics. Genetic antecedents Yuichiro Yano, MD, PhD of factors, such as body mass index and smoking that influence Donald M. Lloyd-Jones, MD, ScM participation, will be subject to a similar bias, which we see in the ALSPAC. Author Affiliations: Department of Preventive Medicine, Northwestern The extent to which bias owing to selection will occur will University Feinberg School of Medicine, Chicago, Illinois (Greenland, Lloyd- depend on the particular association being explored and the op- Jones); Department of Preventive Medicine, University of Mississippi Medical erating selection mechanisms. We suggest that researchers con- Center, Jackson (Yano). sider the potential for selection bias to be affecting their analy- Corresponding Author: Philip Greenland, MD, Department of Preventive Medicine, Northwestern University Feinberg School of Medicine, 680 N Lake ses and carry out sensitivity analyses to assess robustness of their Shore Dr, Ste 1400, Chicago, IL 60611 (p-greenland@northwestern.edu). conclusions to selection bias. Published Online: November 29, 2017. doi:10.1001/jamacardio.2017.4350 Conflict of Interest Disclosures: All authors have completed and submitted the Marcus Munafò, PhD ICMJE Form for Disclosure of Potential Conflicts of Interest and none were George Davey Smith, DSc reported. 1. Yano Y, O’Donnell CJ, Kuller L, et al. Association of coronary artery calcium Author Affiliations: MRC Integrative Epidemiology Unit, University of Bristol, score vs age with cardiovascular risk in older adults: an analysis of pooled Bristol, United Kingdom (Munafò, Smith); UK Centre for Tobacco and Alcohol population-based studies. JAMA Cardiol. 2017;2(9):986-994. Studies, School of Experimental Psychology, University of Bristol, Bristol, United 2. Schenker MP, Dorbala S, Hong EC, et al. Interrelation of coronary Kingdom (Munafò); School of Social and Community Medicine, University of calcification, myocardial ischemia, and outcomes in patients with intermediate Bristol, Bristol, United Kingdom (Smith). likelihood of coronary artery disease: a combined positron emission Corresponding Author: Marcus Munafò, PhD, UK Centre for Tobacco and tomography/computed tomography study. Circulation. 2008;117(13):1693-1700. Alcohol Studies, School of Experimental Psychology, University of Bristol, 12a 3. Greenland P, Bonow RO. How low-risk is a coronary calcium score of zero? Priory Rd, Bristol BS8 1TU, United Kingdom (marcus.munafo@bristol.ac.uk). the importance of conditional probability. Circulation. 2008;117(13):1627-1629. Published Online: December 13, 2017. doi:10.1001/jamacardio.2017.4279 Conflict of Interest Disclosures: Both authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were Biased Estimates in Mendelian Randomization reported. Studies Conducted in Unrepresentative Samples Editorial Note: This letter was shown to the corresponding author of the original article, who declined to reply on behalf of the authors. To the Editor Large cohort studies have transformed our ability 1. Fry A, Littlejohns TJ, Sudlow C, et al. Comparison of sociodemographic and to evaluate complex exposure-outcome relationships. However, health-related characteristics of UK biobank participants with the general these studies are typically not representative of the source popu- population [published online June 21, 2017]. Am J Epidemiol. lation from which they are sampled because of selection at re- 2. Munafò MR, Tilling K, Taylor AE, Evans DM, Davey Smith G. Collider scope: cruitment, attrition over time (in prospective studies), or both. when selection bias can substantially influence observed associations Forexample,inUKBiobank,theresponseratewasapproximately [published online June 7, 2017]. Int J Epidemiol. doi:10.1101/079707 5%, and the resulting sample was healthier and more highly edu- 3. Lyall DM, Celis-Morales C, Ward J, et al. Association of body mass index with cardiometabolic disease in the UK biobank: a Mendelian randomization study. cated than the general population of the United Kingdom. JAMA Cardiol. 2017;2(8):882-889. It is often assumed that while this lack of representativeness 4. Davey Smith G, Ebrahim S. ‘Mendelian randomization’: can genetic is problematic for estimating prevalence, it will make little or no epidemiology contribute to understanding environmental determinants of difference to association analyses. In our view, this is overly op- disease? Int J Epidemiol. 2003;32(1):1-22. timistic. In 2017, we examined the potential effect of selection bias on results obtained from studies with low response rates. We argue that because selection can induce collider bias (which CORRECTION occurs when 2 variables independently influence a third variable Error in Abstract: In the Original Investigation by Pokharel et al titled “Association and that variable is conditioned on), selection can lead to biased ofSerialKansasCityCardiomyopathyQuestionnaireAssessmentsWithDeathandHos- estimates of associations. pitalizationinPatientsWithHeartFailureWithPreservedandReducedEjectionFrac- tion: A Secondary Analysis of 2 Randomized Clinical Trials,” published online Novem- In this context, it is worth considering the analysis in JAMA 3 ber 1, 2017, there was an error in the Results subsection of the Abstract. The first Cardiology by Lyall et al, who reported an association of body sentenceshouldbereplacedwith2sentencesthatread“Of1767eligibleTOPCATpar- mass index with cardiometabolic disease in UK Biobank. The ex- ticipants, 882 were women (49.9%), and the mean (SD) age was 71.5 (9.7) years. Of posure and several of the outcomes (eg, coronary heart disease 2130 eligible HF-ACTION participants, 599 were women (28.1%), and the mean age was 58.6 (12.7) years.” This article was corrected online. and type 2 diabetes) might plausibly be expected to be negatively 1. Pokharel Y, Khariton Y, Tang Y, et al. Association of serial Kansas City associatedwithparticipationinUKBiobank;inwhichcase,aspu- Cardiomyopathy Questionnaire assessments with death and hospitalization in rious negative association between the 2 will be created, and any patients with heart failure with preserved and reduced ejection fraction: true association will be biased, because the use of a sample sub- a secondary analysis of 2 randomized clinical trials [published online November ject to this selection amounts to conditioning on a collider. 1, 2017]. JAMA Cardiol. doi:10.1001/jamacardio.2017.3983 jamacardiology.com (Reprinted) JAMA Cardiology February 2018 Volume 3, Number 2 181 © 2017 American Medical Association. All rights reserved.

Journal

JAMA CardiologyAmerican Medical Association

Published: Feb 13, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off