LETTERS

Oak Street, Vancouver, BC V6H 3N1 Canada (e-mail: [email protected]). Reprints can be ordered at http://www. ajph.org by clicking the “Reprints” link. This letter was accepted April 19, 2014. doi:10.2105/AJPH.2014.302053

Contributors All authors contributed equally to this letter.

References 1. Davis RR, Hofferth SL, Shenassa ED. Gestational weight gain and risk of infant death in the United States. Am J Public Health. 2014;104(suppl 1):S90---S95. 2. Institute of Medicine. Weight Gain During Pregnancy: Reexamining the Guidelines. Washington, DC: National Academies Press; 2009. 3. Whitcomb BW, Schisterman EF, Perkins NJ, Platt RW. Quantification of collider-stratification bias and the birthweight paradox. Paediatr Perinat Epidemiol. 2009;23(5):394---402. 4. Institute of Medicine. Nutrition During Pregnancy. Washington, DC: National Academies Press; 1990. 5. Siega-Riz AM, Adair LS, Hobel CJ. Maternal underweight status and inadequate rate of weight gain during the third trimester of pregnancy increases the risk of preterm delivery. J Nutr. 1996;126(1):146---153. 6. Hutcheon JA, Bodnar LM, Joseph KS, Abrams B, Simhan HN, Platt RW. The bias in current measures of gestational weight gain. Paediatr Perinat Epidemiol. 2012;26:109---116. 7. Hutcheon JA, Platt RW, Abrams B, Himes KP, Simhan HN, Bodnar LM. A weight-gain-for-gestationalage z score chart for the assessment of maternal weight gain in pregnancy. Am J Clin Nutr. 2013;97(5):1062--1067.

DAVIS ET AL. RESPOND We thank Hutcheon et al. for commenting on our article.1 We are in agreement regarding the potential pitfalls of “collider stratification bias,”2 which is conditioning analyses on a variable that is the common effect of multiple causes (e.g., in the case of our article, conditioning on gestational age when assessing the association between gestational weight gain (GWG) and risk of infant mortality). While a discussion of this type of bias is beyond the scope of this communication, we would like to emphasize the importance of subject matter and theoretical considerations when considering the likelihood and extent of such bias. With regard to substantive considerations, Hutcheon et al. assert that at best, adjustment for gestational age only estimates the association between gestational weight gain and infant death if preterm birth were eliminated, and at worst, introduces additional bias. . . .

e2 | Letters

This worst-case scenario is based on directed acyclic graphs models in which a risk factor is assumed to be unconditionally independent of the outcome.2 An assumption of unconditional independence is clearly false in the case of GWG and risk of infant mortality.1,3,4 In other words, this worst-case scenario is predicated on a narrow consideration of what is statistically possible rather than biologically plausible. Moreover, as discussed in the paper, our models of the association between GWG and risk of infant mortality included covariates for two preterm birth categories. As for theoretical considerations, the utility of rigorous methods, such as a weight-gain-forgestational age z-score,5 to control for confounding by gestational age is incontrovertible. However, nationally representative data sets that allow for the analysis of serial GWG do not exist. Rather, the matter of debate is the degree of residual confounding that may result when regression-based methods are used instead.2,6 On another note, consider our key finding that infants born to underweight, normal-weight, and overweight women with inadequate GWG had odds of mortality during infancy that were respectively 6.18, 1.47, and 2.11 times those of infants born to women with adequate GWG. To attribute these distinct effects solely to confounding one must assume differential confounding within strata of BMI, an unlikely proposition. We agree with Hutcheon et al. regarding the utility of their approach, and we also stand by our conclusions. We invite readers to evaluate our conclusions in the context of the above-mentioned substantive and theoretical considerations. j

This letter was accepted May 4, 2014. doi:10.2105/AJPH.2014.302073

Contributors All authors contributed equally to this letter.

References 1. Davis RR, Hofferth SL, Shenassa ED. Gestational weight gain and risk of infant death in the United States. Am J Public Health. 2014;104(suppl 1):S90---S95. 2. Whitcomb BW, Schisterman EF, Perkins NJ, Platt RW. Quantification of collider stratification bias and the birth weight paradox. Paediatr Perinat Epidemiol. 2009;23(5):394---402. 3. Chen A, Feresu S, Fernandez C, Rogan W. Maternal obesity and the risk of infant death in the United States. Epidemiology. 2009;20(1):74---81. 4. Institute of Medicine and National Research Council. Weight Gain During Pregnancy: Reexamining the Guidelines. Washington, DC: National Academy Press; 2009: 1---263. 5. Hutcheon JA, Bodnar LM, Joseph KS, Abrams B, Simhan HN, Platt RW. The bias in current measures of gestational weight gain. Paediatr Perinat Epidemiol. 2012; 26:109---116. 6. Greenland S. Quantifying biases in causal models: classical confounding vs. collider-stratification bias. Epidemiology. 2003;14:300---306.

Regina R. Davis, PhD Sandra L. Hofferth, PhD Edmond D. Shenassa, ScD

About the Authors Regina R. Davis, Sandra L. Hofferth, and Edmond D. Shenassa are with the Department of Family Science, Maternal and Child Health Program, University of Maryland School of Public Health, College Park. Regina R. Davis is also with the American Public Health Association, Washington, DC. Correspondence should be sent to Regina R. Davis, University of Maryland, College Park. Department of Family Science, 1142 School of Public Health Building, College Park, MD 20747 (e-mail: [email protected]). Reprints can be ordered at http://www.ajph.org by clicking the “Reprints” link.

American Journal of Public Health | September 2014, Vol 104, No. 9

Davis et al. respond.

Davis et al. respond. - PDF Download Free
421KB Sizes 0 Downloads 3 Views