Copyright © 2004 by the Johns Hopkins Bloomberg School of Public Health
LETTERS TO THE EDITOR |
RE: "ASSOCIATION OF A WOMAN’S OWN BIRTH WEIGHT WITH HER SUBSEQUENT RISK FOR PREGNANCY-INDUCED HYPERTENSION"
Clinic of Obstetrics, Division of Experimental Obstetrics, Campus Virchow-Klinikum, Charité–University Medicine Berlin, Berlin 13353, Germany
Recently, Innes et al. (1) published findings from a large study on "fetal origins" of pregnancy-induced hypertension (PIH), concluding that a strong inverse relation exists between birth weight and PIH risk. However, it seems to us that important parts of their analysis are seriously biased because of a fundamental conceptual and methodological mistake.
Unadjusted analysis of the data set of Innes et al. (1) demonstrated a U-shaped relation between birth weight and later risk of PIH; that is, both low birth weight and high birth weight appeared to be risk factors for PIH. However, after adjustment for prepregnancy body mass index (weight (kg)/height (m)2), women with high birth weight were no longer at risk of PIH. Consequently, high birth weight was not further considered as a risk factor for PIH, either in the Abstract or in the Discussion section of the paper. Rather, all attention was directed toward low birth weight as a risk factor for later PIH. However, by adjusting for prepregnancy body mass index, the authors obviously controlled for a causal intermediate factor rather than a confounder of the relation between high birth weight and PIH. This can be easily demonstrated using directed acyclic graphs (2, 3). Part A of figure 1 shows the causal structure underlying the analysis, as assumed by Innes et al. Obviously, the implications of this structure are illogical: Prepregnancy body mass index cannot be a confounder of the birth weight-PIH relation, since a woman’s prepregnancy body mass index cannot influence her own birth weight. Rather, two arguments support the idea that the directed acyclic graph shown in part B of figure 1 correctly reflects the causal structure of the relation under investigation here: 1) High birth weight is a risk factor for high body mass index in later life (4). 2) High body mass index is a strong risk factor for PIH (5, 6). Moreover, high body mass index is obviously one of the most important risk factors for PIH in the study by Innes et al. (1). Therefore, high prepregnancy body mass index, a strong risk factor for PIH, is likely to be a descendant of high birth weight or a causal intermediate of the high-birth-weight–PIH relation. As was recently summarized in the Journal (3), adjusting for causal intermediates is seriously discouraged, since it will not remove confounding but will create another biased estimate (2). Therefore, the "adjusted" estimates presented by Innes et al. must be biased. This can be easily demonstrated using figure 1, part B. According to the rules for the use of directed acyclic graphs (2, 3, 7, 8), conditioning on prepregnancy body mass index ([B]) blocks the path between birth weight and PIH. Consequently, birth weight and PIH are conditionally independent. Actually, this is what Innes et al. found: After conditioning on the causal intermediate "prepregnancy body mass index," the relation between high birth weight and PIH disappeared. Moreover, the arguments given above not only imply that Innes et al. cannot conclude that high birth weight is not a risk factor for PIH; they also suggest that the strong association Innes et al. observed between low birth weight and PIH risk after adjustment for prepregnancy body mass index must be a biased estimate as well.
|
We suggest that the estimates presented by Innes et al. (1) for an association between birth weight and PIH risk are biased because of incorrect adjustment on a descendant or causal intermediate factor. Most importantly, however, the report by Innes et al. might lead to incorrect conclusions about the absence of a role of high birth weight as a risk factor for PIH. Since the prevalence of high birth weight has continued to increase in Western industrialized countries during recent years (9), it might be of considerable public health importance to correctly recognize possible long-term consequences of high birth weight.
REFERENCES
- Innes KE, Byers TE, Marshall JA, et al. Association of a woman’s own birth weight with her subsequent risk for pregnancy-induced hypertension. Am J Epidemiol 2003;158:861–70.
[Abstract/Free Full Text] - Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology 1999;10:37–48.[CrossRef][Web of Science][Medline]
- Hernán MA, Hernández-Diaz S, Werler MM, et al. Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology. Am J Epidemiol 2002;155:176–84.
[Abstract/Free Full Text] - Rasmussen F, Johansson M. The relation of weight, length and ponderal index at birth to body mass index and overweight among 18-year-old males in Sweden. Eur J Epidemiol 1998;14:373–80.[CrossRef][Web of Science][Medline]
- Zhang J, Zeisler J, Hatch MC, et al. Epidemiology of pregnancy-induced hypertension. Epidemiol Rev 1997;19:218–32.
[Free Full Text] - Eskenazi B, Fenster L, Sidney S. A multivariate analysis of risk factors for pre-eclampsia. JAMA 1991;266:237–41.
[Abstract/Free Full Text] - Pearl J. Causality. Cambridge, United Kingdom: Cambridge University Press, 2000.
- Robins JM. Data, design, and background knowledge in etiologic inference. Epidemiology 2001;12:313–20.[CrossRef][Web of Science][Medline]
- Bergmann RL, Richter R, Bergmann KE, et al. Secular trends in neonatal macrosomia in Berlin: influences of potential determinants. Paediatr Perinat Epidemiol 2003;17:244–9.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  K. E. Innes, T. E. Byers, J. A. Marshall, and A. Baron FOUR AUTHORS REPLY Am. J. Epidemiol., September 1, 2004; 160(5): 508 - 508. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||