American Journal of Medical Genetics 73:337–344 (1997)

Application of Transmission Disequilibrium Tests

to Nonsyndromic Oral Clefts: Including Candidate
Genes and Environmental Exposures in the Models
Nancy E. Maestri,1,4 Terri H. Beaty,4* Jacqueline Hetmanski,4 E. Anne Smith,3 Iain McIntosh,3,6
Diego F. Wyszynski,4,6 Kung-Yee Liang,5 David L. Duffy,7 and Craig VanderKolk2
1
Department of Pediatrics, School of Medicine, Johns Hopkins University, Baltimore, Maryland
2
Department of Surgery, School of Medicine, Johns Hopkins University, Baltimore, Maryland
3
Center for Medical Genetics, Department of Medicine, School of Medicine, Johns Hopkins University,
Baltimore, Maryland
4
Department of Epidemiology, School of Hygiene and Public Health, Johns Hopkins University, Baltimore, Maryland
5
Department of Biostatistics, School of Hygiene and Public Health, Johns Hopkins University, Baltimore, Maryland
6
Medical Genetics Branch, National Center for Human Genome Research, National Institutes of Health,
Bethesda, Maryland
7
Epidemiology and Population Health Unit, Queensland Institute of Medical Research, Brisbane, Australia

Extensive epidemiological and genetic stud- were incorporated as effect modifiers. We

ies of the cause of oral clefts have demon-
strated strong familial aggregation but have
failed to yield definitive evidence of any
single genetic mechanism. We used the
transmission/disequilibrium test (TDT) to
investigate the relationship between oral
clefts and markers associated with five
candidate genes by utilizing 160 parent-
offspring trios. Conditional logistic regres-
sion models extended the TDT to include co-
variates as effect modifiers, thus permitting
tests for gene-environment interactions.
For four of these candidates [transforming
growth factor alpha (TGFA), transforming
growth factor beta 3 (TGFB3), retinoic acid
receptor (RARA), and the proto-oncogene
BCL3], we detected modestly elevated odds
ratios for the transmission of one marker al-
lele to cleft probands when all the trios were
analyzed together. These odds ratios in-
creased when information on type of cleft,
race, family history, or maternal smoking
detected significant interaction between
maternal smoking and the transmission of
alleles for markers near TGFA and TGFB3;
excess transmission of allele 3 at BCL3 was
most significant among cleft lip probands;
and the odds ratios for transmission of al-
leles at D19S178 and THRA1 were signifi-
cant when ethnic group was included in the
model. We suggest that utilizing an analyti-
cal strategy that allows for stratification of
data and incorporating environmental ef-
fects into a single analysis may be more ef-
fective for detecting genes of small effect.
Am. J. Med. Genet. 73:337–344, 1997.
© 1997 Wiley-Liss, Inc.
KEY WORDS: linkage; birth defects; TGFA;
TGFB3; smoking
INTRODUCTION

Oral clefts are common birth defects, with differing

Abbreviations: BCL3, B-cell Leukemia/lymphoma-3; CP, Cleft
palate; CL, Cleft lip; CLP, Cleft lip and palate; CL/P, Cleft lip
with or without cleft palate; TDT, Transmission/disequilibrium
test; LRT, Likelihood ratio test; CI, Confidence interval; TGFA,
Transforming growth factor alpha; TGFB3, Transforming growth
factor beta 3; RARA, Retinoic acid receptor; THRA1, Thyroid hor-
mone receptor alpha.
Contract grant sponsor: NIH; Contract grant number: RO1-DE-
10293.
*Correspondence to: Dr. Terri H. Beaty, Department of Epide-
miology, Room 6507, School of Hygiene and Public Health, 615
North Wolfe Street, Baltimore, MD 21205-2103.
Received 14 April 1997; Accepted 25 June 1997
© 1997 Wiley-Liss, Inc.
birth prevalence rates across populations [Vanderas,
1987; Sayetta et al., 1989]. Extensive epidemiological
and genetic studies into the cause of oral clefts have
demonstrated strong familial aggregation but have
failed to yield definitive evidence of any single genetic
mechanism [Wyszynski et al., 1996]. Numerous mal-
formation syndromes, which include oral clefts in ad-
dition to structural abnormalities or cognitive deficits,
have been attributed to cytogenetic abnormalities, en-
vironmental factors, or specific gene mutations. Epide-
miological studies have suggested that a proportion of
nonsyndromic oral clefts may be associated with ma-
338 Maestri et al.
ternal exposures during the critical early period of
pregnancy; these exposures include prescription drugs,
cigarette smoking, and alcohol [Wyszynski and Beaty,
1996]. Thus, oral clefts represent a good example of a
common birth defect with a complex and heterogeneous
etiology.
Family studies of nonsyndromic oral clefts suggest
that affected relatives are usually concordant for the
type of cleft, and embryologic studies show the lip and
palate close at different times in development.
Whereas defects in the formation of the primary palate
lead to cleft lip with or without cleft palate during the
sixth week of embryological development, cleft palate
only results from failure of fusion of the medial edges of
the palatal shelves, usually during weeks 8–9. As a
result, genetic studies usually divide clefts into two
anatomical sub-groups: probands with cleft palate only
(CP) and probands with cleft lip with or without cleft
palate (CL/P). To date, segregation analyses of families
ascertained through a proband with either form of oral
clefts have provided inconsistent evidence for major
gene control, multifactorial inheritance, or some mix-
ture of both. Recent data suggest multiple loci may be
involved, i.e. oligogenic control [Farrall and Holder,
1992; Mitchell and Risch, 1992; FitzPatrick and Far-
rall, 1993].
Linkage analyses of multiplex families have sug-
gested linkage of non syndromic cleft lip and palate
(CLP) to markers on chromosomes 6p [Eiberg et al.,
1987; Carinci et al., 1995], 4q [Beiraghi et al., 1994],
and of CL/P to 19q [Stein et al., 1995; Wyszynski et al.,
1997a]. While replication of linkage findings is essen-
tial in studies of complex phenotypes, failure to repli-
cate a reported linkage could be attributed to linkage
heterogeneity, i.e., different loci may control risk in dif-
ferent samples.
Because of the difficulty in ascertaining multiplex
families, population-based association studies have
also been used to investigate the relationship between
oral clefts and markers at specific candidate genes, i.e.,
genes chosen for their potential role in palatogenesis.
Such case-control studies compare the frequency of
marker alleles at a candidate locus in a well-defined
group of (cleft) cases to a similar group of unaffected
controls. Ardinger et al. [1989] first demonstrated such
an association between two RFLPs at the transforming
growth factor alpha (TGFA) locus in a well-defined
group of CL/P cases, suggesting the possibility that an
abnormality in this gene might increase susceptibility
to CL/P. These original findings were replicated in
some population-based studies but not others [see
Wyszynski et al., 1996 for review], and to date no
simple mechanism has been proposed to explain this
association. In particular, all tests for linkage to the
TGFA locus have argued against there being a tradi-
tional Mendelian susceptibility locus near TGFA on
chromosome 2 [Hecht et al., 1991; Vintiner et al., 1992;
Wyszynski et al., 1997b]. Studies suggesting an inter-
action between maternal smoking and TGFA support a
modifying role for the TGFA gene [Hwang et al., 1995;
Shaw et al., 1996]. Mitchell [1997] has recently sug-
gested that conflicting results among populations may
reflect varying proportions of cases with either bilat-
eral CL and/or family history of CL/P.
Evidence for an association between the retinoic acid
receptor (RARA) gene and CL/P was demonstrated in a
Caucasian population [Chenevix-Trench et al., 1992]
and an Asian population [Shaw et al., 1993]. Animal
studies have demonstrated a role for RARA in palato-
genesis [Damm et al., 1993], and retinoic acid is known
to play a role in homeobox gene expression in the de-
veloping head [Studer et al., 1994]. Mitchell et al.
[1995] provided further evidence for an association be-
tween marker alleles on 4q and CL/P, and one extended
family has provided suggestive evidence of linkage to
this region [Beiraghi et al., 1994], but compelling con-
firmatory evidence is still lacking.
Population-based studies may yield false evidence of
an association between a marker allele and a disease
susceptibility locus if confounding due to unrecognized
population stratification exists between the case and
control groups. The transmission/disequilibrium test
(TDT) was developed as an alternative to test explicitly
for segregation distortion due to co-segregation or link-
age [Spielman et al., 1993; Spielman and Ewens, 1996].
By using parent-proband trios, this test statistic com-
pares the number of times a heterozygous parent
transmits a marker allele to an affected child with the
number of times this allele is not transmitted. Feng et
al. [1994] used the TDT to support the association of
the rarer C2 allele at the TGFA locus with susceptibil-
ity to CL/P. Similarly, Wyszynski et al. [1997a] used
the TDT to demonstrate that allele 3 of the proto-
oncogene BCL3 (B-cell leukemia/lymphoma-3) and al-
lele 13 of the marker D19S178 were transmitted more
frequently to CL/P individuals in multiplex families.
Originally, the TDT was designed to detect linkage
and disequilibrium in cases where a population-based
association had already been found; thus, prior knowl-
edge of the allele of interest was necessary. Several
modifications of the TDT have been proposed to extend
this strategy to multiallelic markers [Bickeboller and
Clerget-Darpoux, 1995; Sham and Curtis, 1995;
Kaplan et al., 1997]. Rice et al. [1995] also generalized
the TDT to a multiallelic system and extended the test
to include covariate effects.
The present study was designed to investigate the
relationship between specific candidate genes and oral
clefts using parent-offspring trios. As part of the Mid-
Atlantic Oral Clefts Study [Beaty et al., 1997], we ob-
tained DNA samples from 160 parent-proband trios.
We utilized the TDT to test for linkage between non-
syndromic oral clefts and several candidate genes that
have been implicated in the etiology of oral clefts.
These include TGFA, transforming growth factor
beta-3 (TGFB3), BCL3, RARA, and F13A. Conditional
logistic regression models extended the TDT to include
covariates as effect modifiers, thus permitting tests for
gene-environment interactions.
MATERIALS AND METHODS
Trios
The Mid-Atlantic Oral Clefts Study (MAOCS) ascer-
tained cases of oral clefts born between June 1992, and

July 1996 through treatment centers in Maryland plus
the Craniofacial Clinic of the Children’s National Medi-
cal Center in Washington, D.C. [Beaty et al., 1997].
Secondary ascertainment sources included annual re-
view of records of the Maryland Birth Defects Report-
ing and Information System and referrals from the
Maryland Society for Cleft Lip and Palate Children.
One hundred sixty nonsyndromic oral cleft probands
were available for this study where both parents pro-
vided interview data and a DNA sample. The probands
were examined by a medical geneticist to confirm their
non-syndromic status; DNA sampling from proband
and both parents was obtained either from a venous
blood sample, a blood spot collected on filter paper, or
cheek cells obtained from buccal brushes.
Genotyping
DNA was extracted from blood spots, whole blood, or
buccal brushes and genotyped as described [Wyszynski
et al., 1997a]. The candidate gene markers used here
are presented in Table I. Primer sequences, allele sizes,
and linkage-mapping information are available from
the Genome Data Base (http://gdbwww.gdb.org) for all
markers. Heterozygosity levels were calculated from
observed parental genotypes in this data set.
D2S443 is a tetranucleotide repeat marker located in
the same YAC as the TGFA gene [Wyszynski et al.,
1997b]; the alleles were coded as 1 to 13, from largest to
smallest. D14S61 is an AC repeat marker that is
within 450 kb of TGFB3 [Cruts et al., 1995]. Chromo-
some 19 markers are the same as those reported by
Wyszynski et al. [1997a]. Thyroid Hormone Receptor a
(THRA1) was chosen as the closest dinucleotide repeat
marker for RARA. Recent mapping data localize
THRA1 to 17q11.2–12 and RARA to 17q12 with 0 cM
between them. The alleles were coded from largest (al-
lele 1) to smallest. D6S470 was chosen as a marker for
F13A, located at 6p25.1-p24.3.
Statistical Analysis
The Statistical Package for Social Science (SPSS for
Windows, v. 6.1.2) was used for descriptive statistics.
In order to test for association between each marker
and a susceptibility locus for oral clefts, the multiple-
allelic version of the TDT implemented in SIB-PAIR,
version 0.93 [Duffy,1996; http://www.qimr.edu.au/
davidd.html] was used. McNemar’s test was used to
test for significant excess transmission of each indi-
vidual allele. The TDT statistic calculated by SIB-PAIR
TABLE I. Candidate Genes and Polymorphic Markers Tested in 160
Cleft Trios
Chromosomal
Candidate gene location Marker
TGFA 2p13 D2S443
TGFB3 14q24 D14S61
BCL3 19q13 BCL3
19q13 APOC2 ac1
19q13 APOC2 007
19q13 D19S178
RARA 17q11.2–q12 THRA1
F13A 6q24.2–p23 D6S470
TDT and Oral Clefts 339
is the test of symmetry in the square table of transmit-
ted vs. non-transmitted alleles to each affected child,
based on the Pearson goodness-of-fit test [Duffy et al.,
1995; Bickeboller et al., 1995; Kaplan et al., 1997]; we
will refer to this as the ‘‘global TDT.’’
The PECAN routine of the Epidemiological Graph-
ics, Estimation, and Testing package [EGRET, version
1.02.07 (1995)] was used for conditional logistic regres-
sion analyses. In this analysis, we calculated the odds
of transmission for the allele of interest (identified from
the TDT) vs. its non-transmission. Each parent of the
parent-proband trio generates a matched pair of obser-
vations: the ‘‘case’’ is defined as the transmitted allele,
while the ‘‘control’’ observation is the non-transmitted
allele. In a traditional matched case-control analysis,
such matched pairs are compared for exposure to some
environmental risk factor or other possible covariates.
For the model examined here, the presence of the allele
of interest, i.e., that allele showing significant trans-
mission in the TDT, denoted allele ‘‘Mi’’, can be viewed
as an ‘‘exposure.’’ Thus an indicator variable for ‘‘expo-
sure’’ is coded as ‘‘1’’ if Mi is the transmitted allele, vs.
‘‘0’’ for all other alleles. The logistic regression model is
Log F P~transmission!
1 − P~transmission! G
= a + b1~allele Mi!
where eb1 defines the odds ratio for the transmission of
allele Mi. Since covariate effects specific to the proband
are identical for each matched pair, it is necessary to
create an interaction term to determine if an observ-
able environmental risk factor (RF) or other character-
istic of the proband modifies transmission of allele i.
This term, created by the product of the indicator vari-
able for the allele (Mi) and the indicator variable for the
observed covariate (RF), is coded as ‘‘1’’ if both Mi and
RF are present.
F G
P~transmission!
Log =
1 − P~transmission!
a + b1~allele Mi! + bRF~allele Mi * RF!
In this model, eb1 represents the odds ratio for trans-
mission of Mi to an unexposed proband, and [eb1+bRF] is
the odds ratio for transmission of Mi to an exposed
proband.
Analyses of each candidate gene involved fitting a
nested hierarchy of regression models. Model I tested
only the significance of the odds ratio for excess trans-
mission of the allele of interest. Additional covariates
Of alleles Heterozygosity
12 0.822
16 0.821
6 0.420
14 0.854
14 0.851
19 0.772
17 0.783
14 0.798
340 Maestri et al.

were then added individually, and the likelihood ratio
test (LRT) was used to assess the impact of each co-
variate on the odds of transmitting the allele. Asymp-
totically, this LRT test statistic has a chi-square dis-
tribution, with degrees of freedom equal to the number
of new terms added to the model, and is interpreted as
the change in the odds of transmission among exposed
individuals. The P-value from Wald’s statistic (b/s.d.) is
reported when there is evidence that an individual
odds ratio is significant but the LRT fails to demon-
strate any significant improvement in fit of the ex-
tended model.
RESULTS
The nonsyndromic oral cleft trios include 50 CP pro-
bands (22 males and 28 females), 31 CL probands (20
males and 11 females), and 79 CLP probands (57 males
and 22 females). Eighty-seven percent of the probands
were Caucasian, 7.5% were African-American, and the
remainder were Hispanic or Asian. The 160 trios rep-
resent 153 distinct families, with 7 families contribut-
ing two trios each. However, none of these related trios
constituted double counting of the parents, i.e. there
are no affected full sibs among these 160 affected pro-
bands.
Although 79 (49%) of the 160 trios reported some
family history of oral clefting, there was a considerable
range in the type of affected relatives. Specifically,
among the 79 probands with some family history of oral
clefts, the closest affected relative was first degree for
31 probands (39.2%), second degree for 13 probands
(16.5%), third degree for 10 probands (12.7%), and be-
yond third for 25 probands (31.6%). In the present
study, smoking was the primary environmental risk
factor; 19% of mothers (n 4 29) reported smoking dur-
ing the periconceptual period, i.e. from 3 months before
through 3 months after conception. Family history for
oral clefts (any reported affected relative) and ethnic
group were also considered risk factors.
The TDT was used first to test for significant excess
transmission of alleles from heterozygous parents to
their affected offspring (the proband). Table II shows
results for all cleft trios, stratified by type of cleft. Cleft
lip probands (CL) were initially analyzed separately
from cleft lip and palate probands (CLP). The reported
P-values, based on McNemar’s test statistic, reflect the
probability of transmission of each specific allele ver-
sus its non-transmission. Here we present data for all
alleles which demonstrated excess transmission, i.e.,
with a P-value <0.05, to at least one group of cleft cases.
Conditional logistic regression analysis was then
used to confirm the significant transmission of alleles
identified by the TDT using a three-step strategy. For
each marker, the first model tested for excess trans-
mission of the allele of interest among all cleft cases;
subsequent models then examined transmission within
each cleft group. Finally we tested for the effect of ob-
servable covariates, i.e., maternal smoking, family his-
tory, and ethnic group, in the presence of the allele of
interest (Table III).
For marker D2S443, the TDT results suggested that
allele 4 was transmitted more frequently than expected
among CLP probands (11 transmitted vs. 2 not trans-
mitted, P-value 4 0.013); the global TDT test statistic
was marginally significant (P 4 0.10). A similar excess
transmission was seen among CP probands, although
it was not statistically significant, possibly due to the
small sample size.
Conditional logistic regression analysis confirmed
these results (Table III). The odds ratio for transmis-
sion of allele 4 at marker D2S443 (calculated as eballele 4,
where ballele 4 4 0.734) was 2.08 (95% CI 1.05, 4.15) for
all cleft groups. This odds ratio increased to 5.5 for
transmission to CLP probands (Model II), with signifi-
cant non-transmission to CL probands, consistent with
the TDT results. Although the LRT for Model II com-
pared to Model I was only marginally significant,
Wald’s statistic for each of these cleft types was signifi-
cant at P < 0.05.
Models III and IV incorporated covariates as poten-
tial effect modifiers. Model III shows that there is in-
creased transmission of allele 4 to all offspring of smok-
ing mothers compared to offspring of non-smoking
mothers (OR 4 9.00 vs. 1.46). Similarly, in Model IV
there was increased odds of transmission of allele 4 to
probands in ‘‘simplex’’ families, i.e. with no reported
history of clefting; OR 4 6.50. The small number of
smoking mothers precludes including both covariates
in a single regression model.

TABLE II. TDT Results for all Informative Probands for Each Markera
Cleft palate(CP) Cleft lip (CL) Cleft lip and palate Cleft lip with or without
(50 trios) (31 trios) (CLP) (79 trios) cleft palate (CL/P) (110 trios)
Marker Allele Freq nb T/TNc P-value n T/NT P-value n T/NT P-value n T/NT P-value
D2S443 allele 4 0.07 47 10/5 0.197 28 4/8 0.248 66 11/2 0.013 94 15/10 0.317
D14S61 allele 6 0.32 40 23/10 0.024 27 14/6 0.074 68 34/20 0.057 95 48/26 0.010
BCL3 allele 3 0.74 31 22/15 0.250 20 17/5 0.010* 40 25/17 0.217* 60 42/22 0.012*
D19S178 allele 8d 0.10 38 7/2 0.097 21 7/1 0.034 59 6/13 0.108 80 13/14 0.847
allele 13e 0.44 38 18/20 0.746 21 9/13 0.394 59 41/25 0.049 80 50/38 0.201
THRA1 allele 8 0.34 41 22/16 0.330 23 15/7 0.088** 65 30/21 0.208 88 45/28 0.047*
D6S470 allele 5 0.13 38 7/8 0.796 22 8/4 0.248 62 15/7 0.088 84 23/11 0.040
a
The P-values, based on McNemar’s test statistic, reflect the probability of transmission of each specific allele versus its non-transmission. An * (**)
indicates that the global TDT test statistic, as described in the text, was significant at the 0.05 (0.01) level.
b
Number of informative probands.
c
Ratio of transmitted/non-transmitted alleles.
d
100 bp fragment.
e
86 bp fragment.
 TDT and Oral Clefts 341

TABLE III. Conditional Logistic Regression Analyses Testing for Interaction Between the
Transmission of Specific Alleles and Observable Covariates
Likelihood ratio
Model Description Odds ratio statistic (d.f.) P-value
D2S443 allele 4
I. All trios 2.08 4.667 P 4 0.03
II. CP probands 2.50
CL probands 0.67
CLP probands 5.50 5.253 (2) Model II v. I, P 4 0.07
III. Smoking mothers 9.00
Non-smoking mothers 1.46 3.626 (1) Model III v. I, P 4 0.06
IV. Family history negative 6.50
Family history positive 1.20 4.530 (1) Model IV v. I, P 4 0.03
D14S61 allele 6
I. All trios 1.84 9.625 P 4 0.002
II. CP probands 2.09
CL probands 2.17
CLP probands 1.62 0.446 (2) Model II v. I, P 4 0.80
III. Non-smoking mothers 1.54
Smoking mothers 5.33 4.228 (1) Model III v. I, P 4 0.04
BCL3 allele 3
I. All trios 1.70 7.343 P 4 0.007
II. CP probands 1.50
CL probands 3.40
CLP probands 1.42 2.568 (2) Model II v. I, P 4 0.28
D19S178 allele 8
I. All trios 1.25 0.445 P 4 0.51
II. CP probands 3.00
CL probands 4.00
CLP probands 0.50 7.541 (2) Model II v. I, P 4 0.02
D19S178 allele 13
I. All trios 1.14 0.508 P 4 0.48
II. CP probands 0.85
CL probands 0.71
CLP probands 1.60 3.898 (2) Model II v. I, P 4 0.14
III. Caucasians 5.00
Non-Caucasians 24.98 5.314 (1) Model III v. I, P 4 0.02
THRA1 allele 8
I. All trios 1.44 3.790 P 4 0.05
II. CP probands 1.37
CL probands 2.00
CLP probands 1.32 0.647 (2) Model II v. I, P 4 0.72
III. Non-Caucasian 0.40
Caucasian 1.72 6.028 (1) Model III v. I, P 4 0.01
D6S471 allele 5
I. All trios 1.47 1.533 P 4 0.22
II. CP trios 0.83
CL/P trios 1.82 1.209 (1) Model II v. I, P 4 0.27

The TGFB3 marker D14S61 showed significant
transmission of allele 6 (Table II) to CP probands (23/
10, P 4 0.024) and to CL/P probands (48/26, P 4
0.010). This ratio of transmitted to non-transmitted al-
leles was close to 2:1 for each sub-group of clefts, sug-
gesting that this gene may play some role in the etiol-
ogy of all clefts. When all clefts were considered to-
gether, the excess transmission of allele 6 (71
transmitted vs. 36 not transmitted) was highly signifi-
cant (P 4 0.0007).
Conditional logistic regression analysis confirmed
these results (Table III); the odds ratio for the trans-
mission of allele 6 was 1.84 (95% CI 1.24, 9.36) among
all cleft trios (LRT 4 9.63, 1 df, P 4 0.002). The odds
ratios for each cleft group ranged from 1.62 for CLP
probands to 2.09 for CP probands, and 2.17 for CL pro-
bands. Since there was no evidence that these odds
ratios were statistically different from each other
(Model II vs. Model I, P 4 0.80), the trios were grouped
together to test for significant covariate effects. Model
III shows that the odds ratio increased to 5.3 when
comparing transmission of allele 6 to offspring exposed
to maternal smoking versus those unexposed (LRT 4
4.23, 1 df, p 4 0.04). However, there was no evidence
that this excess transmission was limited to any spe-
cific group of cleft probands, and seemed uniformly in-
creased among all three groups. Neither family history
nor race significantly affected the transmission of
allele 6.
Allele 3 is the most common allele at BCL3, and the
342 Maestri et al.
high proportion of homozygous parents reduces the
number of informative trios by approximately 25%. Ex-
amination of each cleft group (Table II) shows that the
most significant transmission distortion occurred
among CL trios (17/5, P 4 0.01). Although the global
TDT P-values for CL, CLP, and CL/P groups were all
significant (P < 0.05), combining CL and CLP probands
may be inappropriate here, since the ratio of transmit-
ted:nontransmitted alleles in CLP trios was less than
1.5:1. The significant P-value associated with the com-
bined CL/P group is likely due to CL trios alone.
Conditional logistic regression analysis indicated the
odds ratio for transmission of allele 3 was 1.70 (95% CI
1.51, 2.51) among all cleft cases (Table III). Although
the odds ratio increased to 3.40 for transmission to CL
probands, the LRT for heterogeneity of the odds ratios
among the different cleft groups (comparing Model II to
Model I) was not statistically significant. No covariate
had a significant effect on transmission of allele 3.
We investigated three markers closely linked to
BCL3. Wyzynski et al. [1997a] reported excess trans-
mission in multiplex Mexican CLP families of allele 13
at D19S178, which is located 1.1 cM from BCL3 [Mur-
ray et al., 1994]. Classifying by cleft type revealed al-
lele 13 was transmitted more frequently to CLP pro-
bands (41/25, P 4 0.05) in these trios, whereas allele 8
was transmitted more frequently to CP and CL pro-
bands (combined groups, 14 transmitted vs. 3 not
transmitted, P 4 0.008). There was no excess trans-
mission of alleles at either of two markers at APOC2
(which is linked to BCL3 at u 4 2.5 cM; results not
shown).
The differential pattern of excess transmission of
D19S178 alleles 8 and 13 was analyzed by conditional
logistic regression analysis (Table III). The odds ratio
for excess transmission of allele 8 was not statistically
significant among all cleft probands in these condi-
tional logistic models. However, when incorporating
type of cleft into the model, there was evidence of ex-
cess transmission to CL probands (OR 4 4.00), and the
LRT comparing Model II to Model I was statistically
significant (P 4 0.02), suggesting heterogeneity among
types of clefts.
There was a suggestion of excess transmission of
D19S178 allele 13 to CLP probands, although the LRT
comparing Model II to model I was not significant (P 4
0.14). However, when race was included as a covariate
in the model, there was significant transmission of al-
lele 13 among non-Caucasians only (OR 4 24.80, LRT
4 5.34, 1 df, P 4 0.021). The apparent excess trans-
mission of allele 13 to CLP probands was no longer
significantly different from other types of clefts after
race was considered in the model (data not shown).
The TDT showed excess transmission of allele 8 at
THRA1 (Table II) when CL and CLP trios were
grouped together (45/28, P 4 0.047), with CL probands
showing the largest excess transmission (15 alleles
transmitted versus 7 not transmitted; P 4 0.088); the
global TDT P-value for this model was less than 0.001.
Conditional logistic regression analysis (Table III)
showed the odds ratio (1.44, 95% CI 0.99, 2.08)
for transmission of THRA1 allele 8 was marginally
significant among all cleft probands (LRT 4 3.79, 1 df,
P 4 0.052). However, there was no evidence for excess
transmission to CL or CLP probands, and the LRT
comparing Model II to Model I (LRT 4 0.027, 1 df, P 4
0.869) suggests homogeneity among types of clefts. We
did detect evidence of excess transmission of allele 8
among Caucasian probands (Model III, OR 4 1.72),
suggesting that race may be an effect modifier.
The significant excess transmission of allele 5 at lo-
cus D6S470 (Table II) among all CL/P trios (23/11, P 4
0.040) seems to reflect excess transmission among both
CL probands and CLP probands, which individually
failed to reach statistical significance, possibly due lim-
ited sample sizes. However, the odds ratio for trans-
mission of allele 5 at D6S470 (Table III) failed to reach
statistical significance either among all cleft trios
(Model I, P 4 0.22), or among a specific sub-group
(Model II, P 4 0.27), suggesting that the TDT results
seen in Table II may represent a false positive.
DISCUSSION
Nonsyndromic oral clefts are complex birth defects
characterized by an uncertain mode of inheritance, in-
complete penetrance, and heterogeneity both within
and among populations. Oral clefts are also associated
with chromosomal anomalies, single locus Mendelian
syndromes, and teratogenic exposures. Recent data
suggest that genetic susceptibility to clefting may be
the result of several susceptibility loci, acting in a mul-
tiplicative fashion [Farrall and Holder, 1992; Mitchell
and Risch, 1992]. Linkage analyses have suggested
that clefting loci may be located on chromosomes 4q
[Beiraghi et al., 1994], 6p [Eiberg et al., 1987; Carinci
et al., 1995], and 19q [Stein et al., 1995; Wyszynski et
al., 1997a]. Similarly, association studies have sug-
gested that several candidate genes, including TGFA
[Ardinger et al., 1989], TGFB3 [Lidral et al., 1996], and
RARA [Chenevix-Trench et al., 1992] may be involved
in the etiology of oral clefts. We utilized the TDT to
study eight markers associated with five candidate
genes, extending our analyses to include tests for gene-
environment interaction. For four of these candidate
genes, we detected modestly elevated odds ratios for
the transmission of at least one marker allele to cleft
probands when all the trios were analyzed together.
These odds ratios increased when we incorporated in-
formation on type of cleft (i.e., CP, CL, CLP), race, fam-
ily history, or maternal smoking as effect modifiers into
a conditional logistic regression model.
The previously reported association between TGFA
and the rarer allele at the TaqI site among offspring of
smoking mothers was supported by a similar finding
with D2S443, a more polymorphic marker near the
TGFA locus. Whereas the odds ratio for the transmis-
sion of allele 4 was doubled among all cleft trios, it
increased more than five-fold among CLP probands,
more than six-fold among probands with a negative
family history of clefting, and nine-fold among off-
spring of smoking mothers. This finding is consistent
with the hypothesis that a sub-set of cleft cases may be
due to a gene-environment interaction. We propose
that a strong family history of clefting may be due to
the action of gene(s) other than TGFA, and that only in

families without any known family history can the ac-
tion of allele 4 and its interaction with maternal smok-
ing be detected. This would fit with the observation of
Kallen [1997] who reported the effect of maternal
smoking on risk of oral clefts was strongest among
women with no prior history of clefting in their fami-
lies.
There was strong evidence for excess transmission of
D14S61 allele 6 to cleft probands, regardless of the type
of cleft. However, there was also a suggestion of a gene-
environment interaction in these data, since the odds
ratio for transmission of allele 6 increased to 5.3 among
offspring of smoking mothers. Small sample size pre-
cludes analysis of the joint effect and/or interaction of
alleles at both the TGFA and TGFB3 loci and maternal
smoking.
These findings are consistent with the data of Lidral
et al. [1996]. They conducted a mutation screen of
TGFB3 among 60 Caucasian CP cases; they detected
two common variants in exons 1 and 5, and two rare
variants. They also detected linkage disequilibrium be-
tween CL/P and a 58 UTR repeat or the exon 5 variant
among 74 and 50 patients, respectively. Data support-
ing the involvement of TGFB3 in the pathogenesis of
CL/P are of particular interest since mice lacking
TGFB3 demonstrate a cleft palate in the absence of
other craniofacial abnormalities [Kaartinen et al.,
1995; Proetzel et al., 1995]. Kaartinen et al. [1995] sug-
gest that TGFB3 has a unique role in palatal fusion
related to epithelial-mesenchymal interactions.
Stein et al. [1995] reported that a sub-set of CL/P
families showed linkage to BCL3. Wyszynski et al.
[1997a] failed to detect this linkage using likelihood
based LOD score methods but did detect excess trans-
mission of BCL3 allele 3 among both United States and
Mexican multiplex CL/P families. Our analysis of 160
trios (91 informative probands for BCL3) included 36
trios (25 informative probands) from Wyszynski’s
study. Our results confirm this excess transmission of
allele 3 (odds ratio 4 1.7 among all trios), although
these trio data suggest that the effect may be limited to
CL probands. The odds ratio was increased three-fold
among this small group of CL trios, but the P-value was
not statistically significant, possibly reflecting small
sample size. The magnitude of the odds ratio among CL
probands did not change when 12 CL trios from
Wyszynski’s study were excluded from the analysis.
Because BCL3 is not an extremely informative
marker (heterozygosity 4 0.47), Stein et al. [1995] and
Wyszynski et al. [1997a] also examined two highly
polymorphic, closely linked markers. Stein et al. [1995]
found evidence of excess transmission of alleles for
markers D19S178 and APOC2 when all informative
trios were analyzed. Wyszynski et al. [1997a] found
excess transmission of allele 13 at D19S178 only
among trios drawn from the Mexican multiplex fami-
lies (transmitted 19 times, not transmitted five times;
P 4 0.004). Our results were similar; excess transmis-
sion of allele 13 was detected among the relatively
small group of non-Caucasian trios (transmitted ten
times, not transmitted two times, P 4 0.02).
Race was also an effect modifier in the analysis of
THRA1, a marker for the RARA locus on chromosome
TDT and Oral Clefts 343
17. The modestly increased odds ratio (1.44, P 4 0.052)
among all trios increased to 1.71 (P 4 0.014) among
Caucasians. These results are consistent with the re-
port of an association between RARA and CL/P in a
Caucasian population [Chenevix-Trench et al., 1992].
Of interest, the murine clefting locus (clf1) maps to a
region with linkage homology to the RARA interval on
17q [Juriloff and Mah, 1995].
These results support a two-stage strategy for asso-
ciation studies of candidate genes for complex traits.
The initial TDT analyses can be utilized to detect ex-
cess transmission of specific marker alleles near a can-
didate gene. Heterogeneity within the data set may be
detected by including covariates in the conditional lo-
gistic regression models. For example, excess transmis-
sion of a specific marker allele in one ethnic group was
detected by conditional logistic regression analysis pre-
sented in Table III. Gene-environment interaction can
also be detected, as evidenced by the allele-smoking
interaction observed with alleles at D14S61 and
D2S443. The association of different candidate genes
with different types of oral clefts is supported by their
differing roles in primary and secondary palate devel-
opment, i.e., these data suggest that whereas TGFB3
may be important for the correct formation of both the
primary and secondary palate, BCL3 and RARA may
exert their influence predominately on development of
the primary palate.
We suggest that inconsistent results from various
linkage and association studies of oral clefts reflect
both the heterogeneity of the clefting phenotype, i.e.,
multiple susceptibility loci, and variability in environ-
mental risk factors. Utilizing an analytical strategy
that allows for stratification of the data and incorpo-
rating environmental effects into a single analysis may
prove powerful in detecting genes of small effect.
ACKNOWLEDGMENTS
We wish to thank Dr. Susan R. Panny, Department
of Health and Mental Hygiene, Baltimore, MD; Dr.
Cynthia Tifft and Ms. Connie Christion, RN, Depart-
ment of Medical Genetics, Children’s National Medical
Center, Washington, D.C.; Dr. Eric Wulfsberg and
Karen Eanet, M.S., Division of Human Genetics, Uni-
versity of Maryland, Baltimore, MD for their assis-
tance in ascertaining cases.
REFERENCES
Ardinger HH, Buetow KH, Bell GI, Bardach J, VanDemark DR, Murray JC
(1989): Association of genetic variation of the transforming growth fac-
tor alpha gene with cleft lip and palate. Am J Hum Genet 45:348–353.
Beaty TH, Maestri NE, Hetmanski JB, Wyszynski DF, VanderKolk C,
Simpson JC, McIntosh I, Smith EA, Zeiger JS, Raymond GV, Panny
SR, Tifft CJ, Lewanda AF, Cristion CA, Wulfsberg EA. (1997): The
Mid-Atlantic oral clefts study: The genetic epidemiology of oral clefts.
Cleft Palate-Craniofacial J. 34:447–454.
Beiraghi S, Foroud T, Diouhy S, Bixler D, Conneally PM, Delozier-
Blanchet D, Hodes ME (1994): Possible location of a major gene for cleft
lip and palate to 4q. Clin Genet 46:255–256.
Bickeboller H, Clerget-Darpoux F (1995): Statistical properties of the al-
lelic and genotypic transmission/disequilibrium test for multiallelic
markers. Genet Epidemiol 12:865–870.
Carinci F, Pezzetti F, Scapoli L, Padula E, Baciliero U, Curioni C, Tognon
344 Maestri et al.
M (1995): Nonsyndromic cleft lip and palate: Evidence of linkage to a
microsatellite marker on 6p23. Am J Hum Genet 56:337–339.
Chenevix-Trench G, Jones K, Green AC, Duffy DL, Martin NG (1992): Cleft
lip with or without cleft palate: associations with transforming growth
factor alpha and retinoic acid receptor loci. Am J Hum Genet 51:1377–
1385.
Cruts M, Backhovens H, Theuns J, Clark RF, LePaslier D, Weissenbach J,
Goate AM, Martin JJ, Van Broeckhoven C (1995): Genetic and physical
characterization of the early-onset Alzheimer’s disease Ab3 locus on
chromosome 14q24.3. Hum Mol Genet 4:1355–1364.
Damm K, Heyman RA, Umesono K, Evans RM (1993): Functional inhibi-
tion of retinoic acid response by dominant negative retinoic acid recep-
tor mutants. Proc Natl Acad Sci USA 909:2989–2993.
Duffy DL (1995): Screening a 2 cM genetic map for allelic association: A
simulated oligogenic trait. Genet Epidemiol 12:595–600.
Duffy DL (1996): SIB-PAIR, version 0.93 (Computer program). Baltimore.
Eiberg H, Bixler D, Nielsen LS, Conneally PM, John J (1987): Suggestion
of linkage of a major locus for nonsyndromic orofacial cleft with F13A
and tentative assignment to chromosome 6. Clin Genet 32:129–132.
Farrall M, Holder S (1992): Familial recurrence-pattern analysis of cleft lip
with or without cleft palate. Am J Hum Genet 50:270–277.
Feng H, Sassani R, Bartlett SP, Lee A, Hecht J, Malcolm S, Winter RM,
Vintiner GM, Buetow KH, Gasser DL (1994): Evidence from family
studies for linkage disequilibrium between TGFA and a gene for non-
syndromic cleft lip with or without cleft palate. Am J Hum Genet 55:
932–936.
FitzPatrick D, Farrall M (1993): An estimation of the number of suscepti-
bility loci for isolated cleft palate. J Craniofac Genet Dev Biol 13:230–
235.
Hecht JT, Want Y, Blanton SH, Michels VV, Daiger SP (1991): Cleft lip and
palate: No evidence of linkage to transforming growth factor alpha. Am
J Hum Genet 49:682–686.
Hwang S-J, Beaty TH, Panny SR, Street NA, Joseph JM, Gordon S, McIn-
tosh I, Francomano CA (1995): Association study of transforming
growth factor alpha (TGFa) TaqI polymorphism and oral clefts: Indi-
cation of gene-environment interaction in a population-based sample of
infants with birth defects. Am J Epidemiol 141:629–636.
Juriloff DM, Nah DG (1995): The major locus for multifactorial nonsyn-
dromic cleft lip maps to mouse chromosome 11. Mammal Genome 6:
63–69.
Kaartinen V, Voncken JW, Shuler C, Warburton D, Bu D, Heisterkamp N,
Groffen J (1995): Abnormal lung development and cleft palate in mice
lacking TGFb3 indicates defects of epitheliala-mesenchymal interac-
tion. Nat Genet 11:415–421.
Kallen K (1997): Maternal smoking and oral clefts. Cleft Palate-Craniofac
J 34:11–16.
Kaplan NL, Martin ER, Weir BS (1997): Power studies for the transmis-
sion/ disequilibrium tests with multiple alleles. Am J Hum Genet 60:
691–702.
Lidral AC, Leysens N, Basart A, Machida J, Daack-Hirsch S, Leysens M,
Romitti P, Buetow KH, Murray J (1996): Genetic evidence for an etio-
logic role for TGFB3 and MSX1 in nonsyndromic cleft lip and palate.
Am J Hum Genet 59S:A141.
Mitchell LE, Risch N (1992): Mode of inheritance of nonsyndromic cleft lip
with or without cleft palate: A reanalysis. Am J Hum Genet 51:323–
332.
Mitchell LE, Healey SC, Chenevix-Trench G (1995): Evidence for an asso-
ciation between nonsyndromic cleft lip with or without cleft palate and
a gene located on the long arm of chromosome 4. Am J Hum Genet
57:1130–1136.
Mitchell LE (1997): Transforming growth factor a locus and nonsyndromic
cleft lip with or without cleft palate: A reappraisal. Genet Epidem
14:231–240.
Murray JC, Buetow KH, Weber JL, Scherpbier-Heddema T, Manion F,
Quillen J, Scheffield VC, Sunden S, Duyk GM; Genethon: Weissenbach
J, Gyapay G, Dib C, Morrissette J, Lathrop GM, Plaetke R, Odelberg S;
Yale University: Ward D; Centre d’Etude du Polymorphisme Humain
(CEPH): Dausset J, Cohen D, Cann H (1994): A comprehensive human
linkage map with centimorgan density. Cooperative Human Linkage
Center (CHLC). Science 265:2049–2054.
Proetzel G, Pawlowski SA, Wiles MV, Yin MY, Boivin GP, Howles PN, Ding
JX, Ferguson MW, Doetschman T (1995): Transforming growth factor-
beta-3 is required for secondary palate fusion. Nat Genet 11:409–414.
Rice JP, Neuman RJ, Hoshaw SL, Daw EW, Gu C (1995): TDT with co-
variates and genomic screens with mod scores: Their behavior on simu-
lated data. Genet Epidemiol 12: 659–664.
Sayetta RB, Weinrich MC, Coston GN (1989): Incidence and prevalence of
cleft lip and palate: What we think we know. Cleft Palate J 26:242–247.
Sham PC, Curtis D (1995): An extended transmission/disequilibrium test
(TDT) for multi-allele marker loci. Ann Hum Genet 59:323–336.
Shaw D, Ray A, Marazita M, Field L (1993): Further evidence of a rela-
tionship between the retinoic acid receptor alpha locus and nonsyn-
dromic cleft lip with or without cleft palate (CL± P). Am J Hum Genet
53:1156–1157.
Shaw GM, Wasserman CR, Lammer EJ, O’Malley CD, Murray JC, Basart
AM, Tolarova MM (1996): Orofacial clefts, parental cigarette smoking,
and transforming growth factor-alpha gene. Am J Hum Genet 58:551–
561.
Spielman RS, McGinnis RE, Evans WJ (1993): Transmission test for link-
age disequilibrium: The insulin gene region and insulin-dependent dia-
betes mellitus. Am J Hum Genet 57:257–272.
Spielman RS, Ewens WJ (1996): Invited editorial: The TDT and other
family-based tests for linkage disequilibrium. Am J Hum Genet 59:
983–989.
Stein J, Mulliken JB, Stal S, Gasser DL, Malcolm S, Winter R, Blanton SH,
Amos C, Seemanova E, Hecht JT (1995): Nonsyndromic cleft lip with or
without cleft palate: Evidence of linkage to BCL3 in 17 multigenera-
tional families. Am J Hum Genet 57:257–272.
Studer M, Popperl H, Marshall H, Kuroiwa A, Krumlauf R (1994): Role of
a conserved retinoic acid response element in rhombomere restriction
of Hoxb-1. Science 265:1728–1732.
Vanderas AP (1987): Incidence of cleft lip, cleft palate and cleft lip and
palate among races: a review. Cleft Palate J 24:216–225.
Vintiner GM, Holder SE, Winter RM, Malcolm S (1992): No evidence of
linkage between the transforming growth factor-alpha gene in families
with apparently autosomal dominant inheritance of cleft lip and palate.
J Med Genet 29:393–397.
Wyszynski DF, Beaty TH, Maestri NE (1996): Genetics of non-syndromic
oral clefts revisited. Cleft Palate-Craniofacial J 33:406–417.
Wyszynski DF, Beaty TH (1996): Review of the role of potential teratogens
in the origin of human non-syndromic oral clefts. Teratology 53:309–
317.
Wyszynski DF, Maestri NE, McIntosh I, Smith EA, Lewanda AF, Garcia-
Delgado C, Vinageras-Guarneros E, Wulfsberg E, Beaty TH (1997a):
Evidence for an association between markers on chromosome 19q and
nonsyndromic cleft lip with or without cleft palate in two groups of
multiplex families. Hum Genet 99:22–26.
Wyszynski DF, Maestri N, Lewanda AF, McIntosh I, Smith EA, Garcia-
Delgado C, Vinageras-Guarneros E, Wulfsberg E, Beaty TH (1997b):
No evidence of linkage for cleft lip with or without cleft palate to a
marker near the transforming growth factor-alpha (TGFa) locus in two
populations. Hum Hered 47:101–109.