Due 11/27/01 at 1 pm sectionOral clefts, transforming growth factor alpha gene variants, and maternal smoking: a population-based case-control study in Denmark, 1991-1994. American Journal of Epidemiology, Vol 149, Issue 3 248-255Epidemiologic Methods (Epi 203)Problem Set: “Confounding and Interaction II” –Answer KeyDue 11/27/01 at 1 pm section 1. Consider the following abstract: Oral clefts, transforming growth factor alpha gene variants, and maternal smoking: a population-based case-control study in Denmark, 1991-1994. American Journal of Epidemiology, Vol 149, Issue 3 248-255Studies in the United States have indicated that maternal first trimester smoking and infant transforming growth factor alpha (TGFA) locus mutations are associated with non-syndromic cleft lip and/or palate (CLP) and that a synergistic effect of these two risk factors occurs. Based on a Danish case-control study of CLP, the authors studied the effects of smoking and TGFA alleles in an ethnically homogeneous setting. Interview information was obtained for mothers of 302 CLP cases (96% of eligible) and for 567 mothers of nonmalformed children (94% of eligible). Multivariate logistic regression analyses revealed that smoking was associated with a moderately increased risk of cleft lip +/- cleft palate (CL(P)) (odds ratio = 1.40, 95% confidence interval 0.99-2.00). No association between smoking and isolated cleft palate (CP) was observed. TGFA genotype was not associated with either CL(P) or CP, and no synergistic effect with smoking was observed. The "rare" TGFA allele occurred in 25% of both cases and controls compared with an average of 14% in other white control groups. Furthermore, the frequency of CLP in Scandinavia is among the highest in the world. Hence, it is possible that the previously reported association between TGFA and CLP to some degree can be attributable to confounding by ethnicity. a) The authors conclude that the previously reported association between TGFA and cleft lip/palate seen in the US was possibly caused by confounding by ethnicity. What technique did the authors use to prevent confounding by ethnicity in their study? (1 pt)By limiting the study population to the ethnically homogenous population of Denmark, the authors effectively removed any variability in ethnicity and precluded it as a confounder. This technique to preventthe occurrence of confounding is known as restriction (i.e., the study population was restricted to an ethnically homogenous group).b) Using the confounder-exposure-disease triangle drawing illustrated in class, depict how the authors precluded confounding by ethnicity. Specifically, draw which arm or arms are affected. (1 pt)By restricting to an ethnically homogeneous group, the authors precluded any association between ethnicityand the cases/controls sampled for the study. In fact, they also eliminated any association between ethnicity and the exposures as well. By eliminating both of the arms in the confounder-exposure-disease triangle, confounding by ethnicity cannot occur (unless, of course, the population is not as homogeneous as the authors believe, in which case residual confounding by ethnicity may still occur).1Smoking (or TGFA mutation)Cleft lip/palateEthnicityc) In this study, the authors matched controls to cases based on place of birth of the infant (by selecting controls from the same hospital in which the case mother gave birth) and time of birth (by selecting the two most proximate “normal” births in the hospital in which the case mother gave birth). i) Can you speculate as to why place of birth and time were considered as potential confounders? (1 pt)Place of birth and the calendar time during which the gestational period occurred may be markers for environmental exposures that could conceivably cause cleft lip/palate. ii) Are place of birth and time of birth equally as concerning as confounders in the evaluation of the smoking-cleft lip/palate association as they are in the TGFA-cleft lip/palate association? Why or why not? (1 pt)One would think that place and time of birth were more likely to be confounders in the smoking-cleft lip/palate association than in the TGFA-cleft lip/palate association. This is because it is more likely forthere to be an association between place/time of birth and smoking than between place/time of birth and TGFA locus mutations (unless these mutations are themselves caused by environmental agents). d) Comment on the merits of the use of matching (which the authors performed) versus stratification (an alternative strategy used in the analysis phase) to evaluate potential confounding by place of birth and time. (1 pt)Both place and time of birth are complex nominal variables, meaning that they have multiple possible values and there is no order to them. When potential confounders are complex nominal variables, managing them in the analysis phase by stratification becomes very difficult. First, it is possible that for some cases there will not have been a control selected (if controls were randomly selected from the population) who resided in the same district at the same time as the case. This effectively shuts down the stratification procedure. Second, even if there was at least one control selected who happened to live in the same district at the same time as the case, the number of strata needed to conduct the stratification procedure would be huge and the sample size would likely be very small in many of them (and likely quite variable from stratum to stratum) resulting in very imprecise measures of association. (Advanced techniques in the analysis phase such as multivariable regression don’t really solve this latter problem. Complex nominal variables would have to be managed with scores of “indicator” predictor variables in one’s model, and there are only so many such predictor variables that these models can hold.)2. Recently, the effects on public health of economic sanctions in certain countries have been questioned. Consider the following:The effect of economic sanctions on the mortality of Iraqi children prior to the 1991 Persian Gulf War. Am J Public Health 2000 Apr;90(4):546-52OBJECTIVES: This study examined the effect of sanctions on mortality among Iraqi children. METHODS: Theeffects of economic sanctions on health are not well known. Past studies on the effect of economic sanctions on mortality have suffered from unreliable data sources and the collinearity of