Magee-Women's Research Institute Department of Obstetrics, Gynecology, and Reproductive Sciences University of Pittsburgh School of Medicine 204 Craft Avenue Pittsburgh, PA 15213 firstname.lastname@example.org (Bodnar)
Department of Epidemiology University of North Carolina School of Public Health Chapel Hill, North Carolina (Kaufman)
To the Editor:
In a study recently published in Epidemiology, O'Brien and colleagues1 presented results of a metaanalysis examining the relation between prepregnancy body mass index (BMI) and risk of preeclampsia. Quantifying this relation is of major public health importance given the increasing prevalence of obesity, as well as the high risk of maternal and neonatal morbidity and mortality associated with preeclampsia. However, results from this overview should be interpreted with caution because the component studies did not estimate a common causal parameter.
The authors included 13 studies in their metaanalysis. For 4 of these studies,2–5 only unadjusted odds ratios for the BMI–preeclampsia relation were presented. It is likely that unmeasured confounding by sociodemographic variables and health-related behaviors biased these effect measures.
Multivariate analyses were conducted for the remaining 9 studies.6–14 Of these, 26,7 presented results from a causal model (ie, measured confounders were used to adjust the BMI–preeclampsia odds ratio). Three others8–10 had BMI as the primary exposure, but adjusted for variables such as chronic hypertension and gestational diabetes, which are likely on the causal pathway from BMI to preeclampsia.15 The adjusted effect estimates therefore do not represent the total causal effect of BMI, but rather its direct effect, the portion not relayed through these intermediates.16
The final 4 studies using multivariate methods11–14 presented adjusted odds ratios derived from predictive models. The objective of predictive modeling is not to determine the causal effect of an exposure on the outcome, but to best predict the outcome by including all variables associated with it.17 Unlike causal modeling, confounding is not an issue in predictive modeling because there is no “primary exposure.” Including variables that are potentially intermediates on the pathway between a predictor and the outcome is not a problem. Therefore, adjusted odds ratios derived from predictive modeling do not necessarily have causal interpretations and may bias the results of metaanalyses.18
Additionally, O'Brien and colleagues1 included papers that examined either preeclampsia (gestational hypertension and proteinuria) or gestational hypertension alone,1 yet these outcomes are recognized as separate entities19 with different risk factors and clinical findings.20 To assess the causal effect of prepregnancy BMI and hypertensive disorders of pregnancy accurately, subclassification may be preferred.21
1.O'Brien TE, Ray JG, Chan W. Maternal body mass index and the risk of preeclampsia: a systematic overview. Epidemiology
2.Steinfeld JD, Valentine S, Lerer T, et al. Obesity-related complications of pregnancy vary by race. J Matern Fetal Med
3.Bowers D, Cohen W. Obesity and related pregnancy complications in an inner-city clinic. J Perinatol
4.Knuist M, Bonsel GJ, Zondervan HA, et al. Risk factors for preeclampsia in nulliparous women in distinct ethnic groups: a prospective study. Obstet Gynecol
5.Ogunyemi D, Hullett S, Leeper J, et al. Prepregnancy body mass index, weight gain during pregnancy, and perinatal outcome in a rural black population. J Matern Fetal Med
6.Baeten JM, Bukusi EA, Lambe M. Pregnancy complications and outcomes among overweight and obese nulliparous women. Am J Public Health
7.Edwards LE, Hellerstedt WL, Alton IR, et al. Pregnancy complications and birth outcomes in obese and normal weight women: effects of gestational weight change. Obstet Gynecol
8.Thadhani R, Stampfer MJ, Hunter DJ, et al. High body mass index and hypercholesterolemia: risk of hypertensive disorders of pregnancy. Obstet Gynecol
9.Bianco AT, Smilen SW, Davis Y, et al. Pregnancy outcome and weight gain recommendations for the morbidly obese woman. Obstet Gynecol
10.Sebire N, Jolly M, Harris JP, et al. Maternal obesity and pregnancy outcome: a study of 287, 213 pregnancies in London. Int J Obes Relat Metab Disord
11.Ros HS, Cnattingius S, Lipworth L. Comparison of risk factors for preeclampsia and gestational hypertension in a population-based cohort study. Am J Epidemiol
12.Conde-Agudelo A, Belizán J. Risk factors for pre-eclampsia in a large cohort of Latin American and Caribbean women. Br J Obstet Gynecol
13.Sibai BM, Ewell M, Levine RJ, et al. Risk factors associated with preeclampsia in healthy nulliparous women. Am J Obstet Gynecol
14.Lee CJ, Hsieh TT, Chiu TH, et al. Risk factors for pre-eclampsia in an Asian population. Int J Gynecol Obstet
15.Rothman KJ, Greenland S. Precision and validity in epidemiologic studies. In: Rothman KJ, Greenland S, eds. Modern Epidemiology
. Philadelphia: Lippincott-Raven Publishers; 1998:125
16.Robins JM, Greenland S. Identifiability and exchangeability for direct and indirect effects. Epidemiology
17.Clayton D, Hills M. Statistical Models in Epidemiology
. New York: Oxford University Press; 1993;271–281.
18.Greenland S. Meta-analysis. In: Rothman KJ, Greenland S, eds. Modern Epidemiology
. Philadelphia: Lippincott-Raven Publishers; 1998:645
19.Gifford RW, August PA, Cunningham G, et al. Report of the National High Blood Pressure Education Program Working Group on high blood pressure in pregnancy. Am J Obstet Gynecol
20.Lindheimer MD, Roberts JM, Cunningham FG. Chelsey's Hypertensive Disorders in Pregnancy
. Stamford, CT: Appleton & Lange; 1999.
21.Greenland S. Quantitative methods in the review of epidemiologic literature. Epidemiol Rev