American Journal of Epidemiology Copyright by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 53, No. Printed in U.S.A. Susceptibility Loci for Oral Clefts in Denmark Mitchell et al. Evaluation of Two Putative Susceptibility Loci for Oral Clefts in the Danish Population Laura E. Mitchell, Jeffrey C. Murray, Sarah O Brien, and Kaare Christensen 3 Previous studies suggest that the risk of nonsyndromic cleft lip with or without cleft palate (CL±P) and isolated cleft palate (CP) is influenced by genetic variation at several loci and that the relation between specific genetic variants and disease risk may be modified by environmental factors. The present study evaluated potential associations between CL±P and CP and two putative clefting susceptibility loci, MSX and TGFB3, using data from a nationwide case-control study conducted in Denmark from 99 to 99. The potential effects of interactions between these genes and two common environmental exposures, first trimester exposure to maternal cigarette smoke and alcohol intake, were also examined. Analyses of these data provide evidence of an association between the risk of CP and variation at the TGFB3 locus. However, there is no evidence that the risk of CL±P or CP is influenced by gene-environment interactions involving MSX or TGFB3 and either first trimester exposure to maternal cigarette smoke or alcohol consumption. Am J Epidemiol ;53:75. case-control studies; cleft lip; cleft palate; genetics Received for publication March 5,, and accepted for publication September 9,. Abbreviations: CI, confidence interval; CL±P, cleft lip with or without cleft palate; CP, isolated cleft palate; OR, odds ratio. Departments of Biostatistics and Epidemiology and of Pediatrics and Center for Clinical Epidemiology and Biostatistics, The University of Pennsylvania School of Medicine, Philadelphia, PA. Department of Pediatrics, University of Iowa, Iowa City, IA. 3 Center for the Prevention of Congenital Malformations, Institute of Public Health, University of Southern Denmark, Odense, Denmark. Reprint requests to Dr. Laura E. Mitchell, Center for the Prevention of Congenital Malformations, Institute of Public Health, University of Southern Denmark, Sdr. Boulevard 3A, 5 Odense C, Denmark (e-mail: kchristensen@health.sdu.dk). Over the past decade there has been considerable interest in identifying genes that contribute to the development of the nonsyndromic forms of orofacial clefts (i.e., cleft lip with or without cleft palate (CL±P) and isolated cleft palate (CP)), and increasingly there has been an emphasis on determining the role of gene-environment interactions in the development of these conditions. Several potential susceptibility loci for orofacial clefts have been identified, including (but not limited to) TGFA, TGFB3, and MSX (, ). In addition, potential gene-environment interactions involving these three loci and exposure to maternal cigarette smoke and/or alcohol intake during pregnancy have been reported (36). However, the findings for associations between orofacial clefts and specific loci or gene-environment interactions have, in general, not been consistent across studies. There are several explanations for the lack of consistent findings regarding potential genetic risk factors for orofacial clefts. First, analyses of the familial aggregation patterns of CL±P and CP suggest that both of these conditions are likely to be determined by the multiplicative effects of several loci and that the contribution of any one locus, to the overall risk of disease, is likely to be relatively small (7, ). Hence, the sample sizes required to detect such loci are quite large, and failure to detect an association in a given study may simply represent a type II error. Second, it is well recognized that case-control studies may result in spurious disease-marker associations if the case and control samples are drawn from ethnically different groups. It is possible that some of the reported associations between orofacial clefts and putative susceptibility loci result from such uncontrolled confounding and represent type I errors. Finally, common maternal exposures such as smoking and alcohol intake do not appear to be strong risk factors for orofacial clefts (9). Hence, error in the measurement of maternal exposures may quickly obscure the evidence for a true gene-environment interaction. The present study is based on a 3-year, nationwide, casecontrol study conducted in Denmark. There are several aspects of this study that serve to reduce the concerns detailed above. Specifically, the study had a high participation rate and relatively large sample size (CL±P, n = 33; CP, n 3; controls, n 9), cases and controls were ascertained from the same ethnically homogeneous population, and exposure information was collected within a few weeks of birth. Furthermore, exposure information during the first trimester of pregnancy was obtained separately for the periods prior to and after pregnancy recognition. Because some women changed their lifestyle on recognition of pregnancy (), this approach is likely to result in a more accurate assessment of first trimester exposure than one that does not differentiate between these periods. Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3 7
Mitchell et al. MATERIALS AND METHODS This study was based on data obtained during a nationwide, 3-year, case-control study of oral clefts conducted in Denmark (). First trimester maternal exposure information was obtained by interview and from birth records. In addition, infant DNA samples were obtained from stored newborn-screening blood spots. Women who delivered a liveborn child with CL±P or CP and no other major malformation or syndrome, in Denmark, during the period December, 99, to August 3, 99, were eligible to be case mothers. A woman was eligible only if she and the child s father spoke Danish fluently. All 53 hospitals in which births took place in Denmark during this time agreed to participate in the study. Women who were not hospitalized in connection with the birth were excluded from this study. During the study period, only percent of the births in Denmark were home confinements. The mothers of the two preceding births in the hospital where the case mother had delivered were selected as controls. Hence, cases and controls were matched on both time and place of birth. A woman was eligible to serve as a control only if her child had no major malformation and was liveborn and if she and the child s father spoke Danish fluently. If all of these criteria were not met, the mother of the next preceding birth in the hospital was selected. If an eligible woman did not wish to participate in the study, another control was not selected. Infants (cases or controls) with major malformations other than CL±P or CP were not included in the study. Major malformations included anomalies such as neural tube defects, monogenic traits (e.g., Van der Woude), syndromes (e.g., trisomy), and sequences (e.g., Pierre Robin). Anomalies such as clubfoot and syndactylia of the second and third toes were considered minor defects and were not a reason for exclusion from study participation. Further, minimal defects, such as naevi and undescended testes, were not considered as associated malformations. Cases (CL±P or CP) initially suspected to have major associated malformations or syndromes were excluded from the study if the anomalies were confirmed through follow-up in the centralized treatment system. which usually took place weeks after delivery. In contrast, control mothers were usually interviewed in the hospital, days after delivery, at the time of the nurses first visit to the case mother. If a control mother had already been discharged from the hospital, she was interviewed in her home. The study questionnaire was focused on maternal exposures in the first trimester of pregnancy. To determine this period with the greatest accuracy, we used information about the last menstruation s first day and a calendar to identify the dates that delimited the first trimester. Because some women will change their lifestyle on recognition of pregnancy, information concerning first trimester alcohol consumption and smoking was obtained separately for the periods prior to and after pregnancy recognition. Genotyping DNA for case infants and control infants was obtained from newborn screening cards (blood spots on filter paper). These samples are maintained in a biologic bank and were linked to the appropriate study subject by means of a personal identification number that is unique to each Dane. Cases and controls were genotyped for two MSX variants (MSX CA and MSXN) and two TGFB3 variants (TGFB3 CA and TGFB3 X5.). Genotyping for the MSX and the two TGFB3 variants has been reported previously (). The MSXN variant is a newly identified sequence variant, located in the single intron of MSX. This variant was originally detected using a search strategy for single strand conformational variants, and it consists of four alleles. Base numbering (table ) is determined upon complete description of the intron sequence, which has been previously unavailable. Allelic variants are detectable by direct sequencing and by single strand conformational polymorphism variation. To date, allele has been recognized in a single heterozygous individual. Genotypes were read and interpreted blind to the case-control status of samples and were repeated when inadequate or uninterpretable results occurred on the first assay. All samples were independently read by two persons, and discrepancies were resolved by repeat assay or reinterpretation. Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3 Interviews The treatment of orofacial clefts in Denmark is completely centralized and includes an initial hospital visit by a specially trained nurse within the first day(s) after the birth of an affected child. A second home visit, by the same nurse, usually occurs approximately weeks after the initial contact. In the early 99s, this service was maintained by nurses. These nurses acted as interviewers for this study and conducted the interviews in connection with their usual work with the families of children with orofacial clefts. The interviewers were instructed and trained by one of the authors (K. C.) during the fall of 99. A mother was included in the study only if her child was alive at the time planned for the interview. Case mothers were interviewed during their second visit with the study nurse, TABLE. Allele 3 MSXN allelic variants* Nucleotide no. 935 7 9 T C C C * The haplotypes associated with each allele were determined by sequencing of the MSX gene in seven persons (three homozygotes, four heterozygotes). Each nucleotide has a one-letter designation that corresponds to the base in that nucleotide: T, thymine; A, adenine; C, cytosine; G, guanine. A A A G T G T G Am J Epidemiol Vol. 53, No.,
Susceptibility Loci for Oral Clefts in Denmark 9 Statistical analysis Women who were epileptic were excluded from these analyses (two controls, two CL±P, one CP). In addition, one member of each of twin pairs was excluded from these analyses. When members of the twin pair were concordant (i.e., either both affected or both unaffected), one twin was randomly excluded. Among discordant twin pairs, the unaffected twin was excluded, and the affected twin was retained as a case. The sample mean for the time to recognition of pregnancy (39 days) was used for women who were missing data for this variable. Differences in allele frequencies across the four major areas of Denmark (Copenhagen, Seeland, Funen, and Jutland) were examined using Fisher s exact test. In addition, genotype frequencies in controls were examined for departures from Hardy-Weinberg equilibrium using Fisher s exact test. Differences in the genotype frequencies between the cases and controls were evaluated using either Fisher s exact test or χ analysis. When there was evidence that the genotypic frequencies differed in cases and controls, the genotypes were dichotomized into high and low risk categories based on visual inspection of the data, and the strength of the association was measured by the odds ratio and its associated 95 percent confidence interval. Since CL±P and CP are known to have at least partially different etiologies, these and all subsequent analyses were stratified according to cleft type. The association between orofacial clefts and first trimester exposure to maternal cigarette smoke and alcohol consumption was assessed using unconditional logistic regression analysis. In the regression models, exposure to maternal cigarette smoke was categorized either as yes/no or by the number of cigarettes per day (, 9, 9, cigarettes). Similarly, maternal alcohol consumption was categorized either as yes/no or by number of drinks per week (,,, 3, 6, >6 drinks). Potential interactions between maternal first trimester smoking and alcohol consumption and each marker were assessed using the test for homogeneity of the odds ratio described by Breslow and Day (). For the purposes of these analyses, both exposures were dichotomized (yes/no), and Mantel-Haenszel estimates of the common odds ratio were used in the calculation of the test statistic. RESULTS Details of the study sample have been provided elsewhere (). Briefly, interviews were completed for 96 percent of case mothers and 9 percent of control mothers. Controls were interviewed an average of 5 days sooner after birth than were cases (9 vs. days, respectively). Informed consent to use part of the newborn screening card for genotyping was provided for. percent of the case infants and. percent of the control infants. Case mothers and control mothers were similar with respect to maternal age. One of the primary concerns regarding case-control studies of potential genetic risk factors is the possibility that Am J Epidemiol Vol. 53, No., cases and controls are drawn from a genetically heterogeneous population. Specifically, if disease risk and allele frequencies at the marker of interest both vary across the underlying subpopulations, disease-marker associations detected in a case-control study may reflect uncontrolled confounding rather than a causal relation (or linkage disequilibrium between a marker and a disease locus). Although Denmark is ethnically very homogeneous compared with most other countries, there are geographic boundaries within the country that could lead to genetic differences within the Danish population. To determine if such differences were likely to exist, allele frequencies for each of the four markers considered in this study and for the TGFA TaqI polymorphism that was the subject of an earlier report () were estimated separately for controls residing on the two islands (Funen and Seeland) and the peninsula (Jutland) that comprise the majority of the country of Denmark. Controls residing in the Danish capital city of Copenhagen (located on the island of Seeland) were also considered separately, because persons from all areas of Denmark reside in this city. There was no evidence that the allele frequencies for these markers differed across these areas of Denmark (MSX CA, p.57; MSXN, p.6; TGFB3 X5., p.3; TGFB3 CA, p.5; TGFA, p.7), providing further support for the claim that Danes are a relatively homogeneous population. Genotype frequencies for the two MSX markers and the two TGFB3 markers in cases and controls are summarized in table. The genotypic distributions of these markers in controls were not significantly different from the distributions expected under Hardy-Weinberg equilibrium (MSX CA, p.93; MSXN, p.33; TGFB3 X5., p.; TGFB3 CA, p.99). There was no evidence of an association between the MSX CA marker and either CL±P (Fisher s exact test, p.7) or CP (Fisher s exact test, p.6). There was also no evidence of an association between MSXN and CL±P (χ.57, p.75). Considering all observed genotypes at the MSXN marker, there was also no evidence of a statistically significant association between this marker and the risk of CP (χ 3.6, p.). However, the frequency of the 3.3 homozygote was elevated in CP cases (6. percent) relative to that of controls (9. percent). When the 3.3 homozygotes were compared with all other genotypes (. and.3) combined, the p value for the association between this marker and CP decreased but remained nonsignificant (χ.96, p.). The odds ratio for the association of the 3.3 genotype and CP was.6 (95 percent confidence interval (CI):.9,.96). There was no evidence of an association between the TGFB3 X5. marker and either CL±P (Fisher s exact test, p.) or CP (Fisher s exact test, p.6) or between the TGFB3 CA marker and CL±P (χ 5.6, p.9). However, there was evidence of a significant association between the TGFB3 CA marker and CP (Fisher s exact test, p.), and the frequency of the. heterozygote in CP cases (6.3 percent) was markedly higher than that in controls (. percent). When persons carrying at least one copy of allele were combined (i.e., genotypes.,., and.3) Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3
TABLE. Frequency of TGFB3 and MSX genotypes in oral cleft case infants and control infants, Denmark, 9999 Mitchell et al. Controls (n = 73) CL±P* (n = 9) CP* (n = 6) No. MSX CA...3...3...5. 3 6.9 3. 5.9 6.3.. 3.3 3.. % No. % No. % No. % No. % No. % No. % No. % No. % No. % 69 39 7.6 9.7.3 7 5 5.9.6 7. 5.5.5.9 5 5 33. 9.3...5. 7 5 5 5.7 7.6 7. 6 56 7 3.9.3 5. MSX N..3 3.3 3. No. % No. % No. % No. % Controls (n = 5) CL±P (n = 5) CP (n = 59) 9 6.. 3..7 3.6 35.6 95 36 9. 5. 6...5. No. TGFB3 X5.... % No. % No. % Controls (n = ) CL±P (n = 75) CP (n = 59) 36 5 9 5. 5.7 3. 59. 3.7 7...6. Am J Epidemiol Vol. 53, No., Controls (n = ) CL±P (n = 7) CP (n = 56) No. 56 6 5 TGFB3 CA...3..3 % No. % No. % No. % No. % No. % 37. 36. 6. 7 7 36.. 6.3 * CL±P, cleft lip with or without cleft palate; CP, isolated cleft palate. 35.3 6.9. 3 9.. 5. 3.3.3. 3.3.5.6. Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3
Susceptibility Loci for Oral Clefts in Denmark and compared with those with any other genotype, the difference between CP cases and controls became more significant (χ 6.6, p.). The odds ratio for this association (genotype.,., or.3 vs. all other genotypes) was.3 (95 percent CI:.,.9). Consistent with an earlier report from this study (), maternal first trimester smoking was associated with an increased risk of CL±P (odds ratio (OR).3; 95 percent CI:.9,.3) but not with CP (OR.9; 95 percent CI:.56,.5) (table 3). Maternal alcohol consumption was associated with nonsignificant reductions in the risk of both CL±P (OR.75; 95 percent CI:.5,.) and CP (OR.6; 95 percent CI:.5,.) (table 3). Controlling for the effects of exposure to maternal cigarette smoke did not appreciably influence the findings for the association between either CL±P or CP and maternal alcohol consumption (results not presented). There are several possible approaches for assessing the potential effects of interaction between a marker and an exposure on the risk of CL±P or CP. We hypothesized that the effect of each exposure would vary by genotype. However, with the exception of CP and the TGFB3 CA marker, we had no a priori hypotheses regarding which genotypes would be associated with higher exposure-related risks. Because the presence of allele at the TGFB3 CA marker was associated with a significantly increased risk of CP, it was hypothesized that, for infants with at least one copy of this allele (compared with those with zero copies), the risk of CP would be higher in those who were exposed to alcohol or cigarette smoke compared with those who were not. There was no evidence that the relation between first trimester exposure to maternal cigarette smoke or alcohol consumption and either CL±P (table ) or CP (table 5) differed by genotype at any of the four markers. DISCUSSION This study evaluated the relation between genetic variation at two loci (MSX and TGFB3) and two common environmental exposures (first trimester exposure to maternal cigarette smoke and alcohol consumption) and both CL±P and CP. The MSX and TGFB3 loci were selected for evaluation because there are both animal and human data that support the role of these genes in craniofacial development. When knocked out in the mouse, both MSX and TGFB3 generate phenotypes that are largely limited to craniofacial features and include cleft palate (3, ). In addition, a missense mutation in MSX has recently been identified in a family with tooth agenesis and various combinations of CL±P and CP (5). Dental anomalies, which occur more frequently in patients with cleft lip and palate (both within and outside the physical region of the cleft), have also been associated with missense mutations in this gene (6). Further, several association studies have provided evidence that MSX and TGFB3 act as susceptibility loci for CL±P and CP in humans (5, 6, ). This study provided evidence of an association between genetic variation at the TGFB3 locus and CP. Specifically, persons whose genotype included at least one copy of allele of the TGFB3 CA marker were at significantly increased risk of CP compared with persons with other genotypes (OR.3; 95 percent CI:.,.9). Although other TABLE 3. Odds ratios and 95% confidence intervals by the logistic regression model for the association of oral cleft occurrence with smoking and alcohol consumption, Denmark, 9999 Smoking (cigarettes/day) 9 9 Cleft lip with or without cleft palate Isolated cleft palate No. OR* 95% CI* No. OR 95% CI 7 36 63...56.3.7,.6.7,.7.69,.9 6 3 7..9.97.55.7,.7.53,.75.,.3 No. of controls 35 99 5 Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3 Smoking No Yes 7..3.9,.3 6 3..9.56,.5 35 (drinks/week) 3 6 >6 6 6 3 9 5 5..77...3.55.5,.3.5,.3.5,.5.,.5.,.5 9 6..7.7.7.36.6.,..59,.3.3,.99.5, 3.9.5, 3.6 3 75 5 No Yes 6 3..75.5,. 9 5..6.5,. 3 36 * OR, odds ratio; CI, confidence interval. Am J Epidemiol Vol. 53, No.,
Mitchell et al. TABLE. Summary of analyses of gene-environment interactions for cleft lip with or without cleft palate, Denmark, 9999 Marker Exposure Crude exposure OR* Mantel- Haenszel OR 95% CI* Homogeneity X df p value MSX CA.5.69.59.69.,.5.79,...3.. MSXN.3.7.3.7.99,..9,.5.57 3.6.75. TGFB3 X5..3.67.3.67.99,..5,.9.95.6.33. TGFB3 CA.5.75.5.76.6,..5,..6.63.7. TGFB3 CA#.5.7.5.7.5,..5,.9..7..6 * OR, odds ratio; CI, exact confidence interval, unless otherwise noted. Cornfield s 95% confidence interval. Genotype 3. omitted from analysis (n = ; cleft lip with or without cleft palate (CL±P)). Genotype. omitted from analysis (n = ; one control, one CL±P). Genotype 3.3 omitted from analysis (n = 3; two controls, one CL±P). # All genotypes that included at least one allele were combined and compared with all other genotypes. studies have also reported evidence for associations between TGFB3 and orofacial clefts (table 6), this is the first report of an association between isolated CP and this specific TGFB3 marker. It is of note that Romitti et al. (6) did find an association between this TGFB3 CA marker and orofacial clefts. However, in their data, genetic variation at this marker was associated with the risk of CL±P, and risk appeared to be increased in the presence of allele 3 rather than in the presence of allele. This study also provided evidence that suggested an association between CP and MSXN, a new marker located in the intron of this gene. Persons homozyogous for allele 3 of this marker appeared to be at increased risk of isolated CP compared with persons with any other genotype (OR.6; 95 percent CI:.9,.96), but this association was not statistically significant. Lidral et al. () have previously reported associations between CP and the MSX CA marker evaluated in this study. However, this association was not confirmed in our data or in two other case-control studies (6, 7). As previously reported (), we found that first trimester exposure to maternal cigarette smoke is a risk factor for CL±P but not for CP. However, we found no evidence that interactions involving this exposure and either MSX or TGFB3 influenced the risk of CL±P or CP. Maestri et al. (5) reported that the risk of orofacial clefts (CL±P or CP) was Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3 TABLE 5. Summary of analyses of gene-environment interactions for cleft palate, Denmark, 9999 Marker Exposure Crude exposure OR* Mantel- Haenszel OR 95% CI* Homogeneity X df p value MSX CA.9.75.9.7.53,.65.,.7 6.3 6.5.6.59 MSXN.6....6,.53.5,.5 3.9.5..3 TGFB3 X5......6,.9.,..36.3..36 TGFB3 CA.....59,..7,.7.9..3.67 * OR, odds ratio; CI, exact confidence interval, unless otherwise noted. Cornfield s 95% confidence interval. Genotype. omitted from analysis (n = two controls). All genotypes that included at least one allele were combined and compared with all other genotypes. Am J Epidemiol Vol. 53, No.,
Susceptibility Loci for Oral Clefts in Denmark 3 TABLE 6. Summary of published studies of TGFB3 and oral clefts Reference Lidral et al., 997 (7) Maestri et al., 997 (5) Lidral et al., 99 () Romitti et al., 999 (6) Mitchell et al., (this report) Population Philippines Mid-Atlantic Iowa Iowa Denmark significantly influenced by an interaction involving DS6 (a marker located within 5 kilobases of TGFB3) and exposure to maternal cigarette smoke. In addition, Romitti et al. (6) have reported evidence that the risk of both CL±P and CP is influenced by gene-environment interactions involving TGFB3 and exposure to maternal cigarette smoke. Our data do not support an association between first trimester exposure to alcohol and the risk of either CL±P or CP. This finding is consistent with many (9), but not all (), studies. We also found no evidence that the risk of CL±P or CP was influenced by interactions involving this exposure and either MSX or TGFB3. Romitti et al. (6) did, however, report evidence of a statistical interaction between maternal alcohol consumption and allele of the MSX CA marker. In summary, this study provides further evidence of an association between the risk of isolated CP and variation at the TGFB3 locus, as well as evidence suggesting that the risk of this condition may also be influenced by variation at the locus for MSX. However, these data do not confirm previously reported associations between CL±P or CP and specific markers at these loci (e.g., CL±P and allele 3 of the TGFB3 CA marker) nor do they support associations between these conditions and gene-environment interactions involving MSX or TGFB3 and first trimester exposure to either maternal cigarette smoke or alcohol consumption. Such discrepancies may reflect differences in disease etiology across study populations. However, the existence of large differences in the underlying causes of these conditions across study samples seems unlikely, given the similarity of the familial aggregation patterns and epidemiologic characteristics of orofacial clefts across populations (). Differences in design may also account for some of the observed discrepancies across studies of orofacial clefts. The present study is likely to be less biased by confounding due Study design Case-control TDT Case-control TDT Case-control Case-control Marker* DS6 5 UTR. X5. 5 UTR. X5. 5 UTR. X5. X5. Cleft lip with or without cleft palate Cleft palate * Markers: DS6, AC repeat within 5 kilobases of TGFB3 (Cruts et al., 995) (6); 5 UTR., hexanucleotide repeat within 5 nucleotides of transcription initiation site (Lidral et al., 99) (); X5., T C mutation in intron (); (). TDT, transmission disequilibrium test. to ethnic differences between cases and controls than most studies (6, ), because Denmark is ethnically very homogeneous. In addition, the risk of ethnic confounding was further reduced by restricting the sample to parents who spoke fluent Danish. Analysis of allele frequencies for five markers supported the genetic homogeneity of this sample. These analyses indicated that, despite geographic boundaries, the population of Denmark is genetically very homogeneous. Further strengths of this study include a high participation rate (controls, 96 percent; cases, 9 percent) and collection of exposure data within weeks of delivery. This reduces concerns regarding the potentially biasing effects of nonparticipation and recall error. In addition, because some women changed their lifestyles upon recognition of pregnancy, the collection of exposure data separately for the periods prior to and after pregnancy recognition is likely to have reduced the potential for misclassification of exposure status that may occur when only those exposures occurring after pregnancy recognition are evaluated. The present study also had drawbacks. The design made it impossible to blind the interviewers to the case-control status of a participant. However, the interviews were highly structured, and the interviewers were closely monitored throughout the 3-year study period. Another potential limitation of this study was that information on exposure status was obtained by interview after the birth of an affected child and, therefore, subject to recall bias. However, at least for smoking, comparison of these interview data with data obtained from the birth record (which is typically completed prior to delivery) indicated that the interview data are valid and, if anything, cases tended to underreport smoking (). Another limitation of this study was that smoking information was available only for the mother. Paternal smoking has not, however, been reported to be associated with orofacial clefts (). Results Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3 Am J Epidemiol Vol. 53, No.,
Mitchell et al. The most important limitation of this study was that it had limited power to detect gene-environment interactions, particularly for the rarer cleft type (i.e., CP). This limitation is not unique to this study but rather a common feature of studies that seek to identify interactions that contribute to disease etiology. However, the issue of power is of particular concern in studies of genetic markers with multiple (i.e., >) alleles, since there is often no a priori basis for pooling alleles (or genotypes). This is in contrast to many environmental exposures, where exposures categories are easily pooled when individual strata become small (e.g., smoking or alcohol data can be dichotomized into yes/no variables). For markers with multiple alleles and no a priori basis for pooling different genotypes, one is forced to either ) stratify across all possible genotypes, which is likely to produce multiple small cell sizes, or ) evaluate each allele in comparison with all others, which gives rise to multiple comparisons. In either case, the power to detect a significant geneenvironment interaction will be low in a typical study sample. It seems likely that future progress in understanding the contribution of gene-environment interactions to the risk of CL±P and CP will require more powerful approaches for detecting such effects. Different analytical methods, such as the case-series design () and the log-linear approach proposed by Umbach and Weinberg (3), may provide some increase in the power to detect gene-environment interactions, but they do not solve the problems associated with small cell sizes and multiple comparisons. These problems can be overcome only by increasing the overall sample size, which is likely to require multicenter collaborations. To reduced concerns regarding heterogeneity across centers, collaborative efforts will require careful consideration of study design and development of common protocols and diagnostic criteria (). Although meta-analysis of observational data remains controversial, well-designed prospective meta-analyses of CL±P and CP may provide advantages over both retrospective meta-analyses of data from individual studies and studies coordinated through a single center (5). ACKNOWLEDGMENTS This work was supported by Danish grants from the Egmont Foundation Helsefonden (/76-9, /7-9, /6-93), Laegeforeningens Forskingsfond (J.6.5), and the Jacob and Olga Madsens Foundation and by US grants from the National Insititue of Dental and Craniofacial Research (grants R DE 9 and DE 559). The authors express their gratitude to Drs. Jorn Olsen, Bent Norgaard-Pedersen, and Olga Basso for their help in establishing the resources upon which this study is based; to Allison Weber and Bonnie Ludwig for technical assistance; and to the study interviewers: Elisabeth Bayer, Birthe Granhof Black, Anne Marie Frydendall, Jette Gertz, Jytte Gregersen, Jette Moes, Annegrethe Pedersen, Kirsten Plesner (deceased), Marjun Reines, Emma Sakstrup, Lisa Smedegaard, and Vibeke Thinggaard. REFERENCES. Wyszynski DF, Beaty TH, Maestri NE. Genetics of nonsyndromic oral clefts revisited. Cleft Palate Craniofac J 996;33: 67.. Schutte BC, Murray JC. The many faces and factors of orofacial clefts. Hum Mol Genet 999;:539. 3. Hwang SH, Beaty TH, Panny SR, et al. Association study of transforming growth factor alpha (TGFα) TaqI polymorphism and oral clefts: indication of gene-environment interaction in a population-based sample of infants with birth defects. Am J Epidemiol 995;:6936.. Shaw GM, Wasserman CR, Lammer EJ, et al. Orofacial clefts, parental cigarette smoking, and transforming growth factoralpha gene variants. Am J Hum Genet 996;5:556. 5. Maestri NE, Beaty TH, Hetmanski JB, et al. Application of transmission disequilibrium tests to nonsyndromic oral clefts: including candidate genes and environmental exposures in the models. Am J Med Genet 997;73:337. 6. Romitti PA, Lidral AC, Munger RG, et al. Candidate genes for nonsyndromic cleft lip and palate and maternal cigarette smoking and alcohol consumption: evaluation of genotypeenvironment interactions from a population-based case-control study of orofacial clefts. Teratology 999;59:395. 7. Christensen K, Mitchell LE. Familial recurrence-pattern analysis of nonsyndromic isolated cleft palate a Danish registry study. Am J Hum Genet 996;5:9.. Mitchell LE, Christensen K. Analysis of the recurrence patterns for nonsyndromic cleft lip with or without cleft palate in the families of 3,73 Danish probands. Am J Med Genet 996; 6:376. 9. Wyszynski DF, Beaty TH. Review of the role of potential teratogens in the origin of human nonsyndromic oral clefts. Teratology 996;53:397.. 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, 9999. Am J Epidemiol 999;9:55.. Lidral AC, Romitti PA, Basart AM, et al. Association of MSX and TGFB3 with nonsyndromic clefting in humans. Am J Hum Genet 99;63:5576.. Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol. The analysis of case-control studies. Lyon, France: International Agency for Research on Cancer, 9. (IARC scientific publication no. 3). 3. Satokata I, Maas R. MSX-deficient mice exhibit cleft palate and abnormalities of craniofacial and tooth development. Nat Genet 99;6:355.. Proetzel G, Pawlowski SA, Wiles MV, et al. Transforming growth factor-β3 is required for secondary palate fusion. Nat Genet 995;:9. 5. van den Boogaard MH, Dorland M, Beemer FA, et al. MSX mutation is associated with orofacial clefting and tooth agenesis in humans. Nat Genetic ;:33. 6. Vastardis H, Karimbux N, Guthua SW, et al. A human MSX homeodomain missense mutation causes selective tooth agenesis. Nat Genetic 996;3:7. 7. Lidral AC, Murray JC, Buetow KH, et al. Studies of the candidate genes TGFB3, MSX, TGFA, and TGFB3 in the etiology of cleft lip and palate in the Philippines. Cleft Palate J 997; 3:6.. Munger RG, Romitti PA, Daack-Hirsch S, et al. Maternal alcohol use and risk of orofacial cleft birth defects. Teratology 996;5:733. 9. Shaw GM, Lammer EJ. Maternal periconceptional alcohol consumption and risk for orofacial clefts. J Pediatr 999;3: 933.. Lorente C, Cordier S, Goujard J, et al. Tobacco and alcohol use during pregnancy and risk of oral clefts. Am J Public Health ;9:59.. Mitchell LE. Genetic epidemiology of birth defects: nonsyndromic cleft lip and neural tube defects. Epidemiol Rev 997; 9:6. Am J Epidemiol Vol. 53, No., Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3
Susceptibility Loci for Oral Clefts in Denmark 5. Begg CB, Zhang Z. Statistical analysis of molecular epidemiology studies employing case-series. Cancer Epidemiol Biomarkers Prev 99;3:735. 3. Umbach DM, Weinberg CR. Designing and analysing casecontrol studies to exploit independence of genotype and exposure. Stat Med 997;6:733.. Wyszynski DF, Mitchell LE. Report of the newly formed international consortium for oral clefts genetics. Cleft Palate Craniofac J 999;36:7. 5. Margitic SE, Morgan TM, Sager MA, et al. Lessons learned from a prospective meta-analysis. J Am Geriatr Soc 995;3: 359. 6. Cruts M, Backhovens H, Theuns J, et al. Genetic and physical characterization of the early-onset Alzheimer s disease Ab3 locus on chromosome q.3. Hum Mol Genet 995;: 3556. Downloaded from http://aje.oxfordjournals.org/ by guest on June, 3 Am J Epidemiol Vol. 53, No.,