Advertisement

Confounding, causality, and confusion: the role of intermediate variables in interpreting observational studies in obstetrics

  • Cande V. Ananth
    Correspondence
    Corresponding author: Cande V. Ananth, PhD, MPH.
    Affiliations
    Department of Obstetrics and Gynecology, College of Physicians and Surgeons, Columbia University, New York, NY

    Department of Epidemiology, Joseph L. Mailman School of Public Health, Columbia University, New York, NY
    Search for articles by this author
  • Enrique F. Schisterman
    Affiliations
    Epidemiology Branch, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, MD
    Search for articles by this author
Published:April 17, 2017DOI:https://doi.org/10.1016/j.ajog.2017.04.016
      Prospective and retrospective cohorts and case-control studies are some of the most important study designs in epidemiology because, under certain assumptions, they can mimic a randomized trial when done well. These assumptions include, but are not limited to, properly accounting for 2 important sources of bias: confounding and selection bias. While not adjusting the causal association for an intermediate variable will yield an unbiased estimate of the exposure-outcome’s total causal effect, it is often that obstetricians will want to adjust for an intermediate variable to assess if the intermediate is the underlying driver of the association. Such a practice must be weighed in light of the underlying research question and whether such an adjustment is necessary should be carefully considered. Gestational age is, by far, the most commonly encountered variable in obstetrics that is often mislabeled as a confounder when, in fact, it may be an intermediate. If, indeed, gestational age is an intermediate but if mistakenly labeled as a confounding variable and consequently adjusted in an analysis, the conclusions can be unexpected. The implications of this overadjustment of an intermediate as though it were a confounder can render an otherwise persuasive study downright meaningless. This commentary provides an exposition of confounding bias, collider stratification, and selection biases, with applications in obstetrics and perinatal epidemiology.

      Key words

      To read this article in full you will need to make a payment

      Purchase one-time access:

      Academic & Personal: 24 hour online accessCorporate R&D Professionals: 24 hour online access
      One-time access price info
      • For academic or personal research use, select 'Academic and Personal'
      • For corporate R&D use, select 'Corporate R&D Professionals'

      Subscribe:

      Subscribe to American Journal of Obstetrics & Gynecology
      Already a print subscriber? Claim online access
      Already an online subscriber? Sign in
      Institutional Access: Sign in to ScienceDirect

      References

        • Rothman K.J.
        • Lash T.L.
        • Greenland S.
        Modern epidemiology.
        Lippincott Williams & Wilkins, New York (NY)2008
      1. Hernán MA, Robins JM. Causal inference. Boca Raton (FL): Chapman & Hall/CRC, in press.

        • Cole S.R.
        • Hernan M.A.
        Constructing inverse probability weights for marginal structural models.
        Am J Epidemiol. 2008; 168: 656-664
        • Strand K.M.
        • Heimstad R.
        • Iversen A.C.
        • et al.
        Mediators of the association between pre-eclampsia and cerebral palsy: population based cohort study.
        BMJ. 2013; 347: f4089
        • Ananth C.V.
        • Vintzileos A.M.
        Maternal-fetal conditions necessitating a medical intervention resulting in preterm birth.
        Am J Obstet Gynecol. 2006; 195: 1557-1563
        • Basso O.
        • Rasmussen S.
        • Weinberg C.R.
        • Wilcox A.J.
        • Irgens L.M.
        • Skjaerven R.
        Trends in fetal and infant survival following preeclampsia.
        JAMA. 2006; 296: 1357-1362
        • Cheong J.L.
        • Doyle L.W.
        • Burnett A.C.
        • et al.
        Association between moderate and late preterm birth and neurodevelopment and social-emotional development at age 2 years.
        JAMA Pediatr. 2017; : e164805
        • Mann J.R.
        • McDermott S.
        • Griffith M.I.
        • Hardin J.
        • Gregg A.
        Uncovering the complex relationship between pre-eclampsia, preterm birth and cerebral palsy.
        Paediatr Perinat Epidemiol. 2011; 25: 100-110
        • Wu C.S.
        • Nohr E.A.
        • Bech B.H.
        • Vestergaard M.
        • Catov J.M.
        • Olsen J.
        Health of children born to mothers who had preeclampsia: a population-based cohort study.
        Am J Obstet Gynecol. 2009; 201: 269.e1-269.e10
        • Ananth C.V.
        • Basso O.
        Impact of pregnancy-induced hypertension on stillbirth and neonatal mortality.
        Epidemiology. 2010; 21: 118-123
        • Hernandez-Diaz S.
        • Toh S.
        • Cnattingius S.
        Risk of pre-eclampsia in first and subsequent pregnancies: prospective cohort study.
        BMJ. 2009; 338: b2255
        • Mostello D.
        • Kallogjeri D.
        • Tungsiripat R.
        • Leet T.
        Recurrence of preeclampsia: effects of gestational age at delivery of the first pregnancy, body mass index, paternity, and interval between births.
        Am J Obstet Gynecol. 2008; 199: 55.e1-55.e7
        • Greenwood C.
        • Yudkin P.
        • Sellers S.
        • Impey L.
        • Doyle P.
        Why is there a modifying effect of gestational age on risk factors for cerebral palsy?.
        Arch Dis Child Fetal Neonatal Ed. 2005; 90: F141-F146
        • Whitcomb B.W.
        • Schisterman E.F.
        • Perkins N.J.
        • Platt R.W.
        Quantification of collider-stratification bias and the birthweight paradox.
        Paediatr Perinat Epidemiol. 2009; 23: 394-402
        • Hernandez-Diaz S.
        • Wilcox A.J.
        • Schisterman E.F.
        • Hernan M.A.
        From causal diagrams to birth weight-specific curves of infant mortality.
        Eur J Epidemiol. 2008; 23: 163-166
        • Wilcox A.J.
        • Weinberg C.R.
        • Basso O.
        On the pitfalls of adjusting for gestational age at birth.
        Am J Epidemiol. 2011; 174: 1062-1068
        • Schisterman E.F.
        • Whitcomb B.W.
        • Mumford S.L.
        • Platt R.W.
        Z-scores and the birthweight paradox.
        Paediatr Perinat Epidemiol. 2009; 23: 403-413
        • Wilcox A.J.
        Birth weight and perinatal mortality: the effect of maternal smoking.
        Am J Epidemiol. 1993; 137: 1098-1104
        • Wilcox A.J.
        Infant mortality among blacks and whites.
        N Engl J Med. 1992; 327 (author reply 44): 1243
        • Buekens P.
        • Notzon F.
        • Kotelchuck M.
        • Wilcox A.
        Why do Mexican Americans give birth to few low-birth-weight infants?.
        Am J Epidemiol. 2000; 152: 347-351
        • Buekens P.
        • Wilcox A.
        Why do small twins have a lower mortality rate than small singletons?.
        Am J Obstet Gynecol. 1993; 168: 937-941
        • Wilcox A.
        • Skjaerven R.
        • Buekens P.
        • Kiely J.
        Birth weight and perinatal mortality. A comparison of the United States and Norway.
        JAMA. 1995; 273: 709-711
        • Buekens P.
        • Wilcox A.J.
        • Kiely J.
        • Masuy-Stroobant G.
        Birthweight, preterm births and neonatal mortality in Belgium and the United States.
        Paediatr Perinat Epidemiol. 1995; 9: 273-280
        • Lisonkova S.
        • Joseph K.S.
        Left truncation bias as a potential explanation for the protective effect of smoking on preeclampsia.
        Epidemiology. 2015; 26: 436-440
        • Greenland S.
        • Pearl J.
        • Robins J.M.
        Causal diagrams for epidemiologic research.
        Epidemiology. 1999; 10: 37-48
        • Hernan M.A.
        • Hernandez-Diaz S.
        • Robins J.M.
        A structural approach to selection bias.
        Epidemiology. 2004; 15: 615-625
        • Pearl J.
        Causal diagrams for empirical research.
        Biometrika. 1995; 82: 669-710
        • Bandoli G.
        • Palmsten K.
        • Flores K.F.
        • Chambers C.D.
        Constructing causal diagrams for common perinatal outcomes: benefits, limitations and motivating examples with maternal antidepressant use in pregnancy.
        Paediatr Perinat Epidemiol. 2016; 30: 521-528
        • Shrier I.
        • Platt R.W.
        Reducing bias through directed acyclic graphs.
        BMC Med Res Methodol. 2008; 8: 70
        • Schisterman E.F.
        • Cole S.R.
        • Ye A.
        • Platt R.W.
        Accuracy loss due to selection bias in cohort studies with left truncation.
        Paediatr Perinat Epidemiol. 2013; 27: 491-502
        • Textor J.
        • Hardt J.
        • Knuppel S.
        DAGitty: a graphical tool for analyzing causal diagrams.
        Epidemiology. 2011; 22: 745
        • Weinberg C.R.
        Toward a clearer definition of confounding.
        Am J Epidemiol. 1993; 137: 1-8
        • Robins J.M.
        • Morgenstern H.
        The foundations of confounding in epidemiology.
        Machine Modeling. 1987; 14: 869-916
        • Schisterman E.F.
        • Cole S.R.
        • Platt R.W.
        Overadjustment bias and unnecessary adjustment in epidemiologic studies.
        Epidemiology. 2009; 20: 488-495
        • Tronnes H.
        • Wilcox A.J.
        • Lie R.T.
        • Markestad T.
        • Moster D.
        Risk of cerebral palsy in relation to pregnancy disorders and preterm birth: a national cohort study.
        Dev Med Child Neurol. 2014; 56: 779-785
        • Cole S.R.
        • Platt R.W.
        • Schisterman E.F.
        • et al.
        Illustrating bias due to conditioning on a collider.
        Int J Epidemiol. 2010; 39: 417-420
        • Greenland S.
        Quantifying biases in causal models: classical confounding vs collider-stratification bias.
        Epidemiology. 2003; 14: 300-306
        • Kramer M.S.
        • Zhang X.
        • Platt R.W.
        Analyzing risks of adverse pregnancy outcomes.
        Am J Epidemiol. 2014; 179: 361-367
        • Savitz D.A.
        In defense of black box epidemiology.
        Epidemiology. 1994; 5: 550-552
        • Greenland S.
        • Gago-Dominguez M.
        • Castelao J.E.
        The value of risk-factor (“black-box”) epidemiology.
        Epidemiology. 2004; 15: 529-535
        • Vanderweele T.J.
        • Mumford S.L.
        • Schisterman E.F.
        Conditioning on intermediates in perinatal epidemiology.
        Epidemiology. 2012; 23: 1-9
        • Ananth C.V.
        • VanderWeele T.J.
        Placental abruption and perinatal mortality with preterm delivery as a mediator: disentangling direct and indirect effects.
        Am J Epidemiol. 2011; 174: 99-108
        • Mendola P.
        • Mumford S.L.
        • Mannisto T.I.
        • Holston A.
        • Reddy U.M.
        • Laughon S.K.
        Controlled direct effects of preeclampsia on neonatal health after accounting for mediation by preterm birth.
        Epidemiology. 2015; 26: 17-26
        • MacLehose R.F.
        • Kaufman J.S.
        Commentary: the wizard of odds.
        Epidemiology. 2012; 23 (discussion 13-4): 10-12
        • Hinkle S.N.
        • Mitchell E.M.
        • Grantz K.L.
        • Ye A.
        • Schisterman E.F.
        Maternal weight gain during pregnancy: comparing methods to address bias due to length of gestation in epidemiological studies.
        Paediatr Perinat Epidemiol. 2016; 30: 294-304

      Linked Article

      • Reply
        American Journal of Obstetrics & GynecologyVol. 218Issue 3
        • Preview
          We thank Dr Lisonkova for her thoughtful letter,1 in which she raises 3 important issues regarding our article on confounding and intermediate variables.2 First, Dr Lisonkova extols the importance of a properly formulated research question and clinical insights to guide research. We agree and wholeheartedly support her contention. Second, as stated in our article, the directed acyclic graphs (DAG) that we presented are oversimplified for illustration purposes, and Dr Lisonkova takes an issue with this.
        • Full-Text
        • PDF