|
85%
for ECS, and 90% for BT (Table 3); in
the absence of BAER testing these are the
percentages of
affected animals
potentially available for breeding,
and
hence worsening the prevalence of deafness.
Gender differences
in deafness
prevalence were not
seen
in the Dalmatian, BT, or ACD, and differences were not
seen
in the ECS or ES if hearing was
considered
to be a
dichotomous trait (Table 4). The
presence of significant
differences in ES or ECS with the
trichotomous
model for
hearing may reflect an imbalance in
the affected
animals, since the association between
gender
and
deafness lost significance when the trichotomous
model was
replaced by the dichotomous model (ECS p
¼
0:601, ES p ¼ 0:067).
It
is unclear that a trichotomous model
better represents this disorder, since
unilateral
deafness
is logically considered
to be incomplete expression
of deafness that in its complete expression affects
both
ears, and
use of a trichotomous model may insert
additional unjustifiable variance.
One study (Famula et
al., 1996)
suggested that different genes
controlled the
hearing status for each ear; however, this
premise
is not supported by
similar mechanisms for
other
bilateral structures in the body, and the
authors have
since moved to consider other
models for
inheritance of
deafness (Famula et al., 2000).
No
significant gender effect was seen in either the
GMS or the
ESAA data subsets of the ES deafness data,
yet
when
they were combined a highly significant
difference was
seen. The Cochran–Mantel–Haenszel
statistic for
conditional
independence demonstrated this to
be an
example of Simpsons paradox (Agresti,
1996), where
the significance seen
in conditional
associations (the
ES data subsets) is reversed in
marginal associations (subsets
combined).
The non-significant result of the
CMHstatistic
(p ¼ 0:323) showed
that the significance seen
with the combined data
sets was false and
artifactual, and
no significant
gender difference existed. In addition,
significance was not seen
in ES with a
dichotomous model.
These findings may be a reflection of the
significant difference for prevalence
between
the two data
subsets; when the analysis controlled for test
site (GMS
vs ESAA) the gender difference lost
significance.
The
difference in prevalence between the subsets may
be a
result of the fact that submission of BAER
results
to
the ESAA hearing registry is voluntary, which
likely inserts
a sampling bias against inclusion of affected
ES.
Although
several investigators have reported a
significant excess
in deafness in Dalmatian females compared to males
(Greilbrokk,
1994; Holliday et al., 1992; Wood
and Lakhani, 1997, 1998), it is unclear from
a consideration
of possible
genetic mechanisms why such
an
effect might occur. It has been suggested
(Famula et al.,
2001) that these
differences may be reflective
of the
fact that BAER testing is voluntary and
as a result a population
sampling bias may
have been introduced
that
selectively revealed deafness more
frequently in females than
males. Wood, who found a
higher deafness
prevalence
in females (Wood
and Lakhani, 1997), utilized generalized
logistic methods to model
hearing in
1234
Dalmatians in
the UK, simultaneously taking into account
testing site, coat color, gender, parental
hearing,
litter
effects, as
well as interaction effects among all of
the variables. Significant effects were seen for
gender
and
for litter
interaction effects, among others.
It is difficult
to explain why gender effects were seen in that
one
study, but not in this study with more than four
times the
number of animal subjects. It
may be possible
that
founder effects are being seen in the UK or that
relative
geographical restriction effects have had
an impact.
It is also unclear how litter effects and other
variables
might interact with gender to influence
the distribution
of
affected animals beyond what is seen
from
direct prevalence comparisons.
One
other small study
also reported an excess of males affected
(Anderson et
al., 1968). The overall
conclusion
that
must be
drawn
from the findings of this study is that there is
no gender
difference in deafness
prevalence in the
breeds
studied. The eventual identification of the molecular
genetic
cause of
this form of deafness may resolve the
issue
of gender. Dialogue on this
issue will doubtless continue.
Pigmentation
varieties that are not determined by the genes
responsible for white color were not significantly
associated
with deafness. Spot colour in the Dalmatian (black,
liver), roan varieties in ES (orange,
blue, tricolor),
the
two subtypes of parti color in ECS (parti roan vs. parti
white and color), and
the four color varieties in
ACD (blue,
red, blue and tan, blue, black and
tan) showed
no significant association
(Table 5). This outcome was
expected
because the responsible genes –
primarily
the B-b pair (black/liver),
but also the A, C, D, and
E series (Little, 1957)
are not considered
as risk factors
for deafness. However,
color variations resulting from
genes producing white did
show significant
associations
with deafness:
patched Dalmatians were less
likely to be deaf than unpatched, as
reported in
previous
studies (Cattanach,
1999; Famula et al., 2000; Greilbrokk,
1994; Strain et al., 1992),
and white BT were
more
likely to
be deaf than colored BT.
Surprisingly, no
difference was detected between roan
and solid
ECS, but
only one of 60 solid ECS was affected, so the
statistical results
may be uncertain.
In
addition, suppression of iris pigmentation by
white genes
was significantly associated with deafness in
the
Dalmatian,
ES, and ECS (Table 6).
Significance was not seen
in BT, but only one dog of 659 was affected,
again
making
the findings uncertain. Blue
eyes in non-Dalmatian breeds
were rare, but carried a
high association with
deafness when it did occur: two of three
blue-eyed ES
were affected, one of four ECS
was affected, and one of
one BT
was affected. For
comparison, 50.7% of blue-eyed Dalmatians
were affected. In these breeds, the occurrence
of
one
or two blue eyes should suggest a strong
likelihood
that deafness
is present. Significant associations between
blue
eyes and deafness in Dalmatians
have
been reported
in numerous other studies (Cattanach, 1999;
Famula et al., 2000;
Greilbrokk, 1994;
Holliday
et al., 1992;
Muhle et al., 2002; Strain et al., 1992).
Together,
the above combine to reinforce the postulate that
deafness in these breeds is closely linked to
the recessive
alleles of the pigmentation locus S, and that
phenotype
indicators of strong expression of the
gene, such
as blue eyes,
or indicators of weak expression
of the gene,
such as the Dalmatian patch, convey
information on
the likelihood of
deafness.
Studies have shown that deafness
in the Dalmatian has
high heritability, and that the
inheritance is best
modelled as a single major
locus (Famula
et al., 2000; Muhle
et al., 2002). The findings of
this study
of significant
association between deafness
and
parental hearing
status (Table 7) support this. However,
the single major locus
inheritance is not best
modelled as
a simple recessive Mendelian autosome (Famula
et al., 2000; Gaillard et al., 2002;
Juraschko
et al.,
2003; Muhle
et al., 2002), which explains the difficulty
of tracking deafness in pedigrees of affected
animals.
Significant
progress is being made in the identification of genes
responsible for deafness in humans and mice
(Steel
and
Bussoli, 1999; Steel and Kros, 2001).
With
progress being made in sequencing the canine
genome (Ostrander
et
al., 2000) and the recent availability
of a set of
microsatellite markers spanning the canine
genome (Cargill
et al.,
2002; Richman et al., 2001), it is
now
possible to begin whole-genome screens of
DNA from
dogs in pedigrees with
deafness (Cargill et al.,
2001).
Once the gene defect responsible for pigment associated
deafness is
identified,
greater progress in reducing
deafness
prevalence will be possible through
utilization of DNA testing.
|
