Oral clefts are among the most common birth defects.1 The defects result from nonclosure of specific facial structures in weeks 5 through 9 of pregnancy, and require extensive surgical and complementary treatment. For unexplained reasons, Norway has one of the highest recorded prevalences of these defects, particularly of cleft lip (1.5 per 1000 births).2,3
Maternal smoking is an established risk factor for oral clefts. A recent meta-analysis of 24 studies estimated that mothers who smoked during pregnancy had a 1.3-fold increased risk of having a baby with cleft lip with or without cleft palate, and a 1.2-fold risk of cleft palate alone.4 The biologic mechanisms that might underlie this association are unknown. Tobacco smoke contains a large number of toxic chemicals.5,6 Studies of smoking-related cancers have found variations in cancer risk associated with variants of genes that regulate detoxification pathways.7,8 The same detoxification genes may also affect the risk of oral clefts,9–12 modifying the smoking effect on clefting to create interaction between smoking and allelic variants. Relevant detoxification during early pregnancy may occur both in the child and in the mother. The genes of both should therefore be considered in studies of birth outcomes.13
Several candidate genes are related to detoxification of components of cigarette smoke. The arylamine N-acetyltransferases (NAT1 and NAT2) are xenobiotic-metabolizing enzymes that play an important role in the metabolic activation of carcinogenic amines present in cigarette smoke.10,11,14 Cytochrome P450 (CYP1A1) is related to the bioactivation of chemicals such as dioxin in cigarette smoke.15,16 The glutathione S-transferase (GST) enzymes affect the detoxification and secretion of compounds of cigarette smoke.9,11,12,17
We explored the effect of maternal smoking on clefting risk, and the modifying effects of candidate detoxification genes through a population-based case-control, family triad study in Norway.
An overview of the design and recruitment of the study is given in Figure 1. All cases born in Norway 1996–2001 and referred for cleft surgery were ascertained through the 2 surgical departments responsible for all facial clefts repairs in Norway. Affected babies are routinely referred for surgery shortly after birth, at which time the family was invited to participate in the study.
During the same period of time, controls were selected randomly from all live births, with a probability of 4 per 1000 (with minor adjustments during the study), using an automated procedure in the Medical Birth Registry of Norway. Controls were invited by mail through the delivering physician. More details of the design of the study have been published elsewhere.18
Case mothers and fathers were asked to donate both blood samples and cheek swabs. We also asked permission to draw a blood sample from the child during surgery, and to retrieve from the centralized screening laboratory in Norway the left-over portion of the child's sample collected for PKU testing.
In control families, the mother and child provided cheek swab samples, and fathers of babies born after November 1998 (half-way through the study) provided swabs. We also retrieved the PKU samples from control babies.
All participating mothers were asked to complete a questionnaire on a broad spectrum of conditions and exposures. This questionnaire was mailed to participants about 4 months after their delivery. An English translation of this questionnaire can be found at our Web site.19
Closure of the lip occurs around week 5 of embryonic life (before many women are aware of their pregnancy) and is followed by closure of the palate around week 9. We explicitly directed our questions about smoking to the 3 first months of pregnancy. The first questionnaire contained detailed questions on smoking both before pregnancy and during the first 3 months of the pregnancy. Mothers reported average number of cigarettes smoked per day (or per month, if less than 1 per day).
We also asked the mother the average number of hours per day she was within 2 m of a person who was smoking. Passive smoking exposure was defined as the exposure of a nonsmoker to a smoker (within 2 m) for at least 2 hours a day. A categorical dose-response variable was created for smoke exposure, with passive smoking as the lowest level, 1–5 cigarettes a day as the second category, 6–10 cigarettes a day as the third, and 11 or more cigarettes a day as the fourth and highest level of smoking.
The smoking information was tested for reliability against prospectively collected data on smoking from the mother's first prenatal visit (typically around week 10 of pregnancy). Some mothers who smoked earlier in pregnancy might have stopped smoking before the prenatal visit. More important, differential recall between cases and controls should be evident in a comparison of the prenatal report and the postdelivery questionnaire. Evidence of differential recall would indicate response bias, and suggest that prospective smoking information is a more reliable data-source in our analyses.
We obtained surgical records for cases from the hospitals. These records contained information on details of the oral cleft as well as diagnoses of syndromes or other accompanying defects. We also retrieved the Medical Birth Registry record, which contained additional information on birth defects diagnoses, and we asked the mother about diagnoses of the child. A case with any accompanying birth defect or diagnosis of a syndrome reported from any source was categorized as a “nonisolated” case. All other cases were categorized as “isolated.”
DNA was extracted from blood of parents and child in the case group and from cheek swabs from parents and child in the control group. For NAT1 we assayed 2 SNPs with labels T1088A (rs1057126) and C1095A (rs15661). For NAT2 we assayed the SNPs C481T (rs1799929), G590A (rs1799930), and G857A (rs1799931). Two SNPs of CYP1A1 were assayed: M1-T400C (rs4646903) and T1101C (rs1048943) and 2 SNPs of GSTP1: A114V (rs1799811) and A1517G (rs947894). SNP assays were based on a Masscode system (QIAGEN Genomics, Inc, Valencia, CA). The SNPs selected for this study within each gene are believed to have functional effects on enzyme activity and are candidates for biologic effects on risk. We also identified individuals who were homozygous for null-variants of GSTM1 and GSTT1.
Analyses of Maternal Smoking
We tested for a main effect of the exposure (maternal smoking) using a traditional case-control analysis of the 2 case categories: (1) cleft lip with or without cleft palate and (2) cleft palate only. We repeated the analysis for isolated cases only (excluding cases with additional defects). Unconditional logistic regression models in STATA v.9 were used to analyze the case-control data for effects of smoking. Odds ratios (ORs) from these analyses are good estimates of relative risks (RRs) since facial clefts are rare conditions. We estimated both crude ORs and ORs adjusted for mother's education, work status, alcohol intake, folate supplementation and dietary folate, multivitamin supplementation, father's income, and calendar year of baby's birth. We computed 95% confidence intervals (CIs) of these measures of association.
Family triads were the primary basis for genetic analysis. The triad design had the advantage of being immune to effects of population stratification. It also provides more information for haplotype reconstruction, which may compensate for a slightly lower statistical power in single marker analyses.20
All SNPs were first assessed for Hardy-Weinberg equilibrium (and any signs of deviance from Mendelian transmission of alleles) in the control triads after November 1998, when we began to include collection of father's DNA. The absence of data from control fathers in the first half of the study is unrelated to genetic characteristics of the fathers, and thus has no effect on genetic results, other than reducing power.
We restricted our analysis of genetic effects to case types for which smoking had a clear effect. This genetic analysis was based on the case triads.13,21 If the assumptions of Hardy-Weinberg equilibrium holds, we used analyses with a program called Haplin, both for effects of single SNPs and for analyses of haplotypes reconstructed from the SNP markers.21 Haplin is implemented as a package in R.22 The methods used by Haplin are generalizations of other family-based methods such as the TDT and log-linear models.13,23 Estimation was done both for haplotypes carried by the child and for haplotypes carried by the mother. Haplin uses maximum likelihood methods and the EM algorithm. This analytic method allows triads with missing genotype information or missing family members (usually fathers) to be included in the estimation. This assumes that genetic data are missing randomly. Haplin estimates the relative risk for a single dose of a haplotype (heterozygotes combining with any other haplotype) and also for a double dose (homozygotes). The reference category for each estimated haplotype effect is the group not carrying that particular haplotype. When only 2 haplotypes enter the estimation, the reference category is the single category of homozygotes for the common haplotype.
For SNPs or haplotypes associated with clefts in the family-based analyses, we first estimated gene-environment interaction using case-triads. Haplin was used to compare the gene effects between case triads of smoking mothers and nonsmoking mothers. A likelihood-based test for difference in gene effects was calculated.
We also supplemented these analyses with case-control analyses, incorporating smoking and the relevant genetic variants. Although family triads provide some information about haplotypes, unique identification of haplotypes is not possible for every individual. When haplotypes were ambiguous, we estimated probability weights for the alternative haplotypes for the mother and the child using the EM algorithm and maximum-likelihood estimation in Haplin. These probability weights were generated separately for case triads and control triads. Case-control analyses were used with these haplotypes (for both mother and child) to verify the main effects of haplotypes and to estimate interaction with smoking. Case control analyses were performed by logistic regression with probability weights in STATA v.9 (StataCorp, College Station, TX).
We had to use the case-control approach to analyze these genotypes, since standard family-based association analyses require more explicit identification of all genotypes. In a set of supplementary analyses we also tested previously-reported associations with GSTT1, GSTM1, and NAT1.9–12
Table 1 provides descriptive information for mothers and fathers of cases and controls. Based on the retrospective questionnaire, 42% of case mothers and 32% of control mothers reported smoking during the first trimester. This information could be subject to recall bias. To explore this possibility, we compared our smoking information with prospectively-collected information on smoking from the mother's first prenatal visit (at an average of 10 weeks of gestation). Prospective information was missing for 3 cases and 40 controls.
Overall, fewer mothers were recorded as smokers in the prospective information, suggesting either that they had stopped smoking before the prenatal visit or that they underreported smoking at their doctor visit. Except for 1 control, all mothers who had been identified at the prenatal visit as smokers were also identified as smokers by the postbirth questionnaire (Table 2). Among case mothers who reported their smoking in the retrospective questionnaire, 48% reported this only in the questionnaire. This proportion was virtually the same for control mothers (49%). The biggest difference between the 2 data sources was for mothers smoking 1 to 5 cigarettes per day. Even so, the proportion of smokers in this category added by the questionnaire was very similar for mothers of cases and controls (64% vs. 62%).
Thus, there was no evidence of differential recall of smoking by cases and controls in our questionnaire. Because only half of the women who reported retrospectively that they smoked in the first trimester also reported at their prenatal visit that they were current smokers, we concluded that the prospective information did not adequately capture first-trimester smoking. We have therefore used the questionnaire information on smoking in our primary analyses.
Effects of Maternal Smoking
There was little evidence of an effect of smoking on the risk of cleft palate only. When we restricted to isolated cleft palate (Table 3), the test-for-trend P value was 0.74. Use of the prospective information on smoking did not alter this (P for trend = 0.37).
In contrast, there was a strong and consistent dose-response effect of smoking for cleft lip (with or without cleft palate), including a small increased risk with passive smoking. Restricting to 314 isolated cases of cleft lip, the risk ranged from 1.6-fold for passive smoking (adjusted OR = 1.59 [95% CI = 1.02–2.47]) to almost 2-fold when mothers smoked more than 10 cigarettes per day ([0.92–4.01]) (Table 4, test-for-trend P value = 0.001). The estimated associations were similar when we used smoking recorded at first prenatal visit (test-for-trend P value = 0.04). Among the cleft lip cases without other defects (“isolated”) there was little evidence of an effect of smoking (overall P value = 0.54; data not shown). Extrapolating from these population-based data, we estimate that 19% of isolated cleft lip cases in Norway may be attributable to maternal smoking in the first trimester.
Family Triad Analyses of Smoking Detoxification Genes
Single Nucleotide Polymorphisms
The call rate of our SNP-assays varied between 88% and 96% (Table 5). None of the SNPs in our study had significant deviance from Hardy-Weinberg equilibrium or Mendelian transmission of alleles among control triads. There was strong linkage disequilibrium among SNPs within each gene in our data.
Since our analysis of smoking showed a clear association only for isolated cleft lip, we limited our analyses of smoking detoxification genes to this case group. There was some evidence of an effect of the NAT2 SNP G590A/rs1799930 (Table 5). Children heterozygous for the rare variant appeared to have a 2-fold risk (RR = 2.0 [95% CI = 1.2–3.5]). The rare variant CYP1A1 SNP T1101C/rs1048943 (3% allele frequency) was estimated with high risks for homozygotes, although only 2 cases were homozygous for this variant. No other SNPs—whether carried by the mother or the infant—appeared to be associated with cleft lip.
The numbers of haplotypes that could be reconstructed from the SNPs (and that were prevalent enough to be included in the estimation) ranged from 2 for NAT1 and CYP1A1 to 3 for GSTP1 and 4 for NAT2 (Fig. 2).
The only haplotype associated with isolated cleft lip was the C-A-G haplotype of NAT2 (referring to the variants of the SNPs rs1799929, rs1799930, and rs1799931, respectively) when this haplotype was carried by the child. This haplotype carries the rare A-allele of G590A/rs1799930 and the common alleles of rs1799929 and rs1799931. A single copy of this haplotype increased the risk of isolated cleft lip 1.6-fold (RR = 1.60 [CI = 1.10–2.40]) and a double dose of the haplotype increased the risk 2.5-fold (2.50 [1.40–4.60]). The effect of haplotypes carrying the very rare variant of CYP1A1 T1101C/rs1048943 could not be estimated.
Using haplotype information derived from the family triads, we then used a case-control approach to estimate haplotype effects for the C-A-G haplotype of NAT2. This did not confirm our family-based analysis. For the child's haplotypes, there was no increased risk for the C-A-G haplotype in single dose (OR = 1.0 [CI = 0.8–1.4]) and only a 1.2-fold risk in double dose (1.2 [0.8–1.9]).
Gene-Environment Interaction Analyses
We explored interaction by creating a dichotomous variable of maternal smoking exposure. In case-triad analyses with Haplin, the association of the C-A-G haplotype of NAT2 was not substantially different among children with smoking mothers (RR = 1.5 in single dose and 2.7 in double dose) and children with nonsmoking mothers (RR = 1.8 in single dose and 2.3 in double dose) (overall P for difference = 0.20). Similarly, there was no evidence of effect modification by smoking for the single SNPs NAT2 rs1799930 and CYP1A1 rs1048943 (overall P-values of 0.93 and 0.81, respectively).
There was no indication of interaction between the child's NAT2 C-A-G haplotype and maternal smoking (P for interaction = 0.72). For the single SNP NAT2 rs1799930 we found no evidence of association in case-control analyses (P = 0.37) and no evidence of interaction with smoking (P = 0.15). The variant of CYP1A1 rs1048943 was too rare to be studied in case-control analyses. Results were again similar when the prospective information on smoking was used.
We found no evidence of association of the null variants of GSTT1 or GSTM1 with isolated cleft lip (Table 6) or of interaction between GSTT1 null in the child and maternal smoking (P for interaction = 0.60). The risk was not changed when both mother and child had the GSTT1 null variant (OR = 0.9 [CI = 0.4–2.2]). When both mother and child had the GSTM1 null variant, however, the risk was 2-fold (2.2 [1.1–4.4]). When restricting to mothers exposed to tobacco smoke, risks were not elevated for children with the NAT1 1088 AA genotype compared with children with TT (1.3 [0.5–3.2]) of for children with the NAT1 1095 AA genotype compared with CC (1.3 [0.6–2.7]). There was no increased risk of cleft lip for children with null variants of both GSTT1 and GSTM1 when the mother smoked (1.0 [0.3–3.2]).
We found persuasive evidence of an association between mothers’ smoking and the risk of cleft lip in her offspring, but no evidence that genetic variation in several detoxification enzymes affected or modified this risk.
This effect of maternal smoking on cleft lip is consistent with several reports in the literature.3,4,24–28 For cleft palate, past evidence is less consistent. Some studies25,28,29 have found increased risk for cleft palate with smoking, but not all.26,27 While we cannot exclude the possibility of an effect of maternal smoking on cleft palate only based on our data, the evidence is much weaker than for cleft lip.
We found an effect of passive smoking on cleft lip among nonsmoking mothers when passive smoking was defined as having a smoking person closer than 2 m for at least 2 hours per day. Passive smoking—defined in a variety of ways—has been studied previously. Shaw et al25 defined passive smoking by closeness to a smoker, and found increased risk among offspring of exposed women. No associations were found in studies defining exposure based on duration of exposure and degree of smokiness in the air,28 or as any exposure to smoke for a nonsmoking mother.30
Our case-control study of smoking has several strengths. It is population-based, with virtually 100% ascertainment of clinically verified cases during a defined time period. Data collection was done in the first months after birth, which should have reduced recall problems for smoking and other exposures. Participation was reasonably good for both cases (88%) and controls (76%). Since the difference between prospective and retrospective report of smoking was nearly identical for cases and controls, we may assume that reporting bias of smoking was minimal. Our data suggest that only half the mothers who smoked in the first trimester reported this at their first antenatal visit, either because they had stopped smoking when they became aware of their pregnancy or because they did not want to tell their doctor. We also had extensive data on relevant confounders for statistical adjustment of our estimates.
The metabolism genes studied here have been of interest in relation to cancer and interactions with smoking because of their activating/deactivating activities for carcinogens. Since mutation caused by the same carcinogens is a possible mechanism for oral clefts, these genes are also relevant candidate genes to explain an effect of smoking on oral clefts.
Since the effect of smoking in our data seemed to be restricted to isolated cleft lip, we limited our genetic analyses for the detoxification genes to this case group. We found some evidence of an effect of NAT2 in the case-family-based analysis, although this was not confirmed in the case-control analysis. The NAT2 haplotype associated with cleft lip carries the A-allele of the G590A-SNP (rs1799930). This variant is known to reduce expression and stability of NAT2 immunoreactive protein and reduce acetylation activity,31,32 and has previously been found to be associated with cleft lip.11 There was, however, no evidence in our data that smoking modified an effect of the NAT2-haplotype on risk of cleft lip. NAT2 may still have a real effect on cleft lip, but apparently not by interaction with smoking.
We also found an indication in our case-triad analyses of an effect of CYP1A1 T1101C/rs1048943, apparently as a strong dominant effect of the rare allele. The association was created by only a few case families and could not be verified in case-control analyses. Larger studies would be needed to study an involvement of this variant with the smoking effect. Other associations previously described for children with variants of NAT1, GSTT1, and GSTM1 for cleft lip were not replicated in our data.9–12
Combining 2 Study Designs
The genetic aspects of our study were strengthened by being able to combine a family-triad design with a case-control design.33 In studies with genetic data for both case-parent triads and control-parent triads, the integrated use of the 2 analytic approaches has several advantages. First, the 2 structures of genetic analysis compensate for each other's weakness. Genetic case-control analyses can be vulnerable to admixture problems in the population, even in relatively homogeneous populations such as Norwegian.34,35 Case-triad analyses overcome this difficulty, but in turn may be vulnerable to deviance from Mendelian transmission of alleles. Control-parent triads allow a check for Mendelian transmission. Second, the integrated approach provides a check for consistency between the effects estimated by case-triad and case-control analyses. A failure to find consistency between the 2 approaches (as occurred with our NAT2 analysis) raises doubts about associations that otherwise might be regarded as strong in a single approach.
Although our overall participation was good, participation was lower for controls than for cases, which creates the possibility of differential participation and bias. The apparent specificity of an effect of smoking to 1 case-group may indicate that such bias did not affect our results, but the possibility cannot be ruled out.
While we validated our smoking information against prospectively collected information, we cannot rule out incompleteness in our information on smoking. If some mothers who smoked had been mistakenly categorized as exposed only to passive smoking, this could bias the passive smoking finding.
We did not have DNA for genetic analyses for control fathers recruited before November 1998. This created a substantial number of missing fathers in our data. Furthermore, our study used DNA from blood cells for case-triads and buccal cells for control triads. If these 2 sources of DNA gave different genotyping results, our case-control analyses of genetic markers could be biased. There is however little evidence for such differences between buccal cell and blood cell DNA.36,37 To assess the probability of genotyping errors more generally, we performed a blinded second genotyping of a random 10% of our samples, which showed an over-all concordance rate of 99.4%. Concordance was not different for cases and controls. Most of the errors were for the null-variant assays.
In sum, we found strong evidence that maternal smoking in the first trimester causes cleft lip. There was some evidence that a functional variant of NAT2 is associated with cleft lip, independent of smoking, although we could not confirm this in our case-control analysis. In our exploration of genes involved with metabolism of cigarette smoke toxicants, we were not able to demonstrate interactions of smoking with NAT1, NAT2, CYP1A1, GSTP1, GSTT1, or GSTM1. Previously reported associations were not confirmed.
1.Christensen K. Methodological issues in epidemiological studies of oral clefts. In: Wyszynski DF, ed. Cleft Lip and Palate From Origin to Treatment
. Oxford: Oxford University Press; 2002:101–107.
2.Harville EW, Wilcox AJ, Lie RT, Vindenes H, Abyholm F. Cleft lip and palate versus cleft lip only: are they distinct defects? Am J Epidemiol
3.Little J, Mossey P. Epidemiology of oral clefts: an international perspective. In: Wyszynski DF, ed. Cleft Lip and Palate From Origin to Treatment
. Oxford: Oxford University Press; 2002:127–158.
4.Little J, Cardy A, Munger RG. Tobacco smoking and oral clefts: a meta-analysis. Bull World Health Organ
5.Baron JA, Rohan TE. Tobacco. In: Schottenfield D, Fraumeni JF, eds. Cancer Epidemiology and Prevention
. New York: Oxford University Press; 1996:269–289.
6.International Agency for Research on Cancer. IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans. Vol. 38. Tobacco Smoking
. Switzerland: International Agency for Research on Cancer; 1986.
7.Vineis P. The relationship between polymorphisms of xenobiotic metabolizing enzymes and susceptibility to cancer. Toxicology
8.Hecht SS. Cigarette smoking and lung cancer: chemical mechanisms and approaches to prevention. Lancet Oncol
9.van Rooij IA, Wegerif MJ, Roelofs HM, et al. Smoking, genetic polymorphisms in biotransformation enzymes, and nonsyndromic oral clefting: a gene-environment interaction. Epidemiology
10.Lammer EJ, Shaw GM, Iovannisci DM, Van Waes J, Finnell RH. Maternal smoking and the risk of orofacial clefts: susceptibility with NAT1 and NAT2 polymorphisms. Epidemiology
11.Shi M, Christensen K, Weinberg CR, et al. Orofacial cleft risk is increased with maternal smoking and specific detoxification-gene variants. Am J Hum Genet
12.Lammer EJ, Shaw GM, Iovannisci DM, Finnell RH. Maternal smoking, genetic variation of glutathione S-transferases, and risk for orofacial clefts. Epidemiology
13.Wilcox AJ, Weinberg CR, Lie RT. Distinguishing the effects of maternal and offspring genes through studies of “case-parent triads.”Am J Epidemiol
14.Sim E, Payton M, Noble M, Minchin R. An update on genetic, structural and functional studies of arylamine N-acetyltransferases in eucaryotes and procaryotes. Hum Mol Genet
15.Muto H, Takizawa Y. Dioxins in cigarette-smoke. Arch Environ Health
16.Dragin N, Dalton TP, Miller ML, Shertzer HG, Nebert DW. For dioxin-induced birth defects, mouse or human CYP1A2 in maternal liver protects whereas mouse CYP1A1 and CYP1B1 are inconsequential. J Biol Chem
17.Wormhoudt LW, Commandeur JN, Vermeulen NP. Genetic polymorphisms of human N-acetyltransferase, cytochrome P450, glutathione-S-transferase, and epoxide hydrolase enzymes: relevance to xenobiotic metabolism and toxicity. Crit Rev Toxicol
18.Wilcox AJ, Lie RT, Solvoll K, et al. Folic acid supplements and risk of facial clefts: national population based case-control study. BMJ
20.Morton NE, Collins A. Tests and estimates of allelic association in complex inheritance. Proc Natl Acad Sci USA
21.Gjessing HK, Lie RT. Case-parent triads: estimating single- and double-dose effects of fetal and maternal disease gene haplotypes. Ann Hum Genet
22.Haplin. Software for analyzing case-parent triad (trio) data with SNP haplotypes. Available at: www.uib.no/smis/gjessing/genetics/software/haplin/
23.Weinberg CR, Wilcox AJ, Lie RT. A log-linear approach to case-parent-triad data: assessing effects of disease genes that act either directly or through maternal effects and that may be subject to parental imprinting. Am J Hum Genet
24.Khoury MJ, Gomez-Farias M, Mulinare J. Does maternal cigarette smoking during pregnancy cause cleft lip and palate in offspring? Am J Dis Child
25.Shaw GM, Wasserman CR, Lammer EJ, et al. Orofacial clefts, parental cigarette smoking, and transforming growth factor-alpha gene variants. Am J Hum Genet
26.Christensen K, Olsen J, Norgaard-Pedersen B, et al. Oral clefts, transforming growth factor alpha gene variants, and maternal smoking: a population-based case-control study in Denmark, 1991–1994. Am J Epidemiol
27.Lieff S, Olshan AF, Werler M, et al. Maternal cigarette smoking during pregnancy and risk of oral clefts in newborns. Am J Epidemiol
28.Little J, Cardy A, Arslan MT, Gilmour M, Mossey PA. Smoking and orofacial clefts: a United Kingdom-based case-control study. Cleft Palate Craniofac J
29.Meyer KA, Williams P, Hernandez-Diaz S, Cnattingius S. Smoking and the risk of oral clefts: exploring the impact of study designs. Epidemiology
30.Honein MA, Rasmussen SA, Reefhuis J, et al. Maternal smoking and environmental tobacco smoke exposure and the risk of orofacial clefts. Epidemiology
31.Fretland AJ, Leff MA, Doll MA, Hein DW. Functional characterization of human N-acetyltransferase 2 (NAT2) single nucleotide polymorphisms. Pharmacogenetics
32.Hein DW, Fretland AJ, Doll MA. Effects of single nucleotide polymorphisms in human N-acetyltransferase 2 on metabolic activation (O-acetylation) of heterocyclic amine carcinogens. Int J Cancer
33.Weinberg CR, Umbach DM. A hybrid design for studying genetic influences on risk of diseases with onset early in life. Am J Hum Genet
34.Thomas DC, Witte JS. Point: population stratification: a problem for case-control studies of candidate-gene associations? Cancer Epidemiol Biomarkers Prev
35.Wacholder S, Rothman N, Caporaso N. Counterpoint: bias from population stratification is not a major threat to the validity of conclusions from epidemiological studies of common polymorphisms and cancer. Cancer Epidemiol Biomarkers Prev
36.Feigelson HS, Rodriguez C, Welch R, et al. Successful genome-wide scan in paired blood and buccal samples. Cancer Epidemiol Biomarkers Prev
37.Weiss JR, Baer MR, Ambrosone CB, et al. Concordance of pharmacogenetic polymorphisms in tumor and germ line DNA in adult patients with acute myeloid leukemia. Cancer Epidemiol Biomarkers Prev