healthmedicinet

The screening results for all Dachshunds radiographically screened for IDC in Finland
until 1 May 2015 were retrieved from the Finnish Dachshund Club open database and
the Finnish Kennel Club breeding database. Screening was initiated in 1998 as a project
of the Finnish Dachshund Club. Since June 2013, it has been possible to record the
screening result in the Finnish Kennel Club’s open breeding database. In both databases,
the screening results are recorded as the number of calcified discs (NCD).

The screening data included the registration number of the dog, date of screening
and the screening result (NCD). The total number of dogs screened was 1567. Fourteen
dogs were foreign and were removed, because they were not included in the pedigree
data. The NCD varied from 0 to 21 (Table 1). The proportion of dogs screened out of all dogs registered was 4.9 % in 1999, 5.7 %
during 2000–2004, 8.0 % during 2005–2009 and 4.6 % during 2010–2013.

Table 1. Distribution of dogs by the number of calcified discs (NCD) in 1553 Dachshunds

The same veterinarian evaluated all the radiographs. The date of screening varied
between 24 November 1998 and 28 April 2015. The dogs were of six different breed variants:
standard and miniature wirehaired, standard and miniature smooth-haired, and standard
and miniature longhaired variants. In all, 1369 of the dogs screened had at least
three generations of pedigree information in the pedigree data.

Pedigree information was obtained from the Finnish Kennel Club. The total number of
dogs in the data was 53,606. The data included the registration number, breed variant
number, sex, year of birth of the dog, as well as the registration number of the dam
and the sire.

Genetic links are apparent between the breed variant populations. Dachshunds have
always been transferred between breed variants according to their phenotypes, and
transfers between miniature and standard variants and from wirehaired to smooth-haired
variants are common. Since 2010, breeding between variants has been allowed, with
the exception of that between longhaired and wirehaired variants. The cross-variant
dog’s size is registered according to the larger parent, and coat according to the
phenotype. Both the size and coat-type can be changed later to match the true phenotype.
During the years 2010–2014, the proportions of these cross-variant dogs out of all
the Dachshunds born in Finland were 2.5 % in the standard wirehaired and 6.8 % in
the miniature wirehaired variants, 2.3 % in the standard smooth-haired and 4.6 % in
the miniature smooth-haired variants, and 6.8 % in the standard longhaired and 0.7 %
in the miniature longhaired variants.

The age at the time of the radiographic screening was calculated from the difference
between the date of screening and the date of birth. The dogs were grouped based on
their age: 24, 24–30, 31–36, 37–42, 43–48 and 48 months. The effect of age on the
NCD was examined with the F statistic in an analysis of variance (ANOVA) type III.
The other candidate effects included in the genetic model were also tested. The model
used was

where NCD
_ijklm
 = number of calcified discs, µ = the overall mean, age
_i
 = fixed effect of the ith age class (i = 1–6), sex
_j
 = fixed effect of the jth sex (j = 1–2), year
_k
 = fixed effect of the kth screening year (k = 1–5, i.e., 2005, 2005–2007, 2008–2010,
2011–2013 and 2013), variant
_l
 = fixed effect of the lth breed variant (l = 1–6) and ?
_ijklm
 = a random residual effect. The residuals were assumed to be independent and ? ~ N(0,
?2). Information on the date of screening was lacking for 112 dogs, and the ANOVA
thus included only 1441 dogs (Table 2). P-values ?0.05 were considered statistically significant.

Table 2. Least-squares (LS) means of the effects in the analysis of variance (ANOVA type III)
for the number of calcified discs (NCD) in 1441 Dachshunds

Since the distribution of the NCDs was highly skewed (Table 1), a log-transformed NCD (lnNCD) was also used as an analysed trait. Transformation
was performed, using the formula lnNCD = ln(1 + NCD). The model described above was
also used for this trait in the ANOVA analysis.

To estimate the heritabilities for the NCD and lnNCD, the variance components were
estimated by restricted maximum likelihood (REML), using REML variance component estimation
(VCE4) 17]. Estimation was done with the following model:

where NCD
_ijklmn
 = number of calcified discs, lnNCD
_ijklmn
 = log-transformed NCD, µ = the overall mean, sex
_i
 = fixed effect of the ith sex (i = 1–2), year
_j
 = fixed effect of the jth screening year (j = 1–5, i.e., 2005, 2005–2007, 2008–2010,
2011–2013, and 2013), variant
_k
 = fixed effect of the kth breed variant (k = 1–6), ec
_l
 = random effect of the lth litter, a
_m
 = random additive genetic effect of the mth animal and ?
_ijklmn
 = a random residual effect. The distributions of ec, a and ? were assumed to be multivariate
normal with zero means and with Var(ec) = I?
_ec²
, Var(a) = A?
_a²
and Var(?) = I?
_?²
, where I and A are the identity matrix and the numerator relationship matrix, respectively.
Dogs born on the same day to the same parents were assumed to form a litter. The number
of litters was 1104. The number of dogs screened per litter varied between one and
four. Despite many litters having only one screened dog, the effect of the litter
was included in the model, since omitting it would probably have resulted in overestimation
of the heritability.

Treating all the subpopulations (breed variants) as one in the genetic analyses may
have led to small biases in the genetic variance estimate. The number of dogs per
variant was too small to perform separate analyses. The differences between the subpopulations
were accounted for by including the breed variant in the model.

The genetic trends for the NCD and lnNCD were assessed, using best linear unbiased
prediction (BLUP) breeding values, which were estimated with the same animal model
as the heritabilities, using the program Multivariate Prediction and Estimation (PEST)
18]. The mean EBV of the dogs born during 1997–2000 was set at 100 and the standard deviation
at 10. After standardization, the scale of the EBVs was reversed so that EBVs larger
than 100 indicated better-than-average breeding values. The genetic trends were assessed
by comparing the mean EBVs and the phenotypic trend by comparing the mean NCD of dogs
born in different years.

The pedigree in the genetic analyses included a total of 9027 dogs, i.e., all the
individuals with a screening result, as well as those behind them in their pedigree.
In all, 2297 of the dogs were base animals with unknown parents. The genetic analyses
also allowed missing effects, i.e., the dog was included in the analysis, even if
it did not have information recorded on all the effects included in the model. Thus,
the number of dogs with screening results in the analyses was 1553.