Estimation of heritabilities, genetic correlations, and breeding values of four traits that collectively define hip dysplasia in dogs.
ABSTRACT OBJECTIVE-To estimate heritabilities and genetic correlations among 4 traits of hip joints (distraction index [DI], dorsolateral subluxation [DLS] score, Norberg angle [NA], and extended-hip joint radiograph [EHR] score) and to derive the breeding values for these traits in dogs. ANIMALS-2,716 dogs of 17 breeds (1,551 dogs in which at least 1 hip joint trait was measured). PROCEDURES-The NA was measured, and an EHR score was assigned. Hip joint radiographs were obtained from some dogs to allow calculation of the DI and DLS score. Heritabilities, genetic correlations, and breeding values among the DI, DLS score, NA, and EHR score were calculated by use of a set of multiple-trait, derivative-free, restricted maximum likelihood computer programs. RESULTS-Among 2,716 dogs, 1,411 (52%) had an estimated inbreeding coefficient of 0%; the remaining dogs had a mean inbreeding coefficient of 6.21%. Estimated heritabilities were 0.61, 0.54, 0.73, and 0.76 for the DI, DLS score, NA, and EHR score, respectively. The EHR score was highly genetically correlated with the NA (r = -0.89) and was moderately genetically correlated with the DI (r = 0.69) and DLS score (r = -0.70). The NA was moderately genetically correlated with the DI (r = -0.69) and DLS score (r = 0.58). Genetic correlation between the DI and DLS score was high (r = -0.91). CONCLUSIONS AND CLINICAL RELEVANCE-Establishment of a selection index that makes use of breeding values jointly estimated from the DI, DLS score, NA, and EHR score should enhance breeding programs to reduce the incidence of hip dysplasia in dogs.
- SourceAvailable from: Zhiwu Zhang[Show abstract] [Hide abstract]
ABSTRACT: Genome-Wide Association Studies shed light on the identification of genes underlying human diseases and agriculturally important traits. This potential has been shadowed by false positive findings. The Mixed Linear Model (MLM) method is flexible enough to simultaneously incorporate population structure and cryptic relationships to reduce false positives. However, its intensive computational burden is prohibitive in practice, especially for large samples. The newly developed algorithm, FaST-LMM, solved the computational problem, but requires that the number of SNPs be less than the number of individuals to derive a rank-reduced relationship. This restriction potentially leads to less statistical power when compared to using all SNPs. We developed a method to extract a small subset of SNPs and use them in FaST-LMM. This method not only retains the computational advantage of FaST-LMM, but also remarkably increases statistical power even when compared to using the entire set of SNPs. We named the method SUPER (Settlement of MLM Under Progressively Exclusive Relationship) and made it available within an implementation of the GAPIT software package.PLoS ONE 01/2014; 9(9):e107684. · 3.53 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Background The inheritance of most human diseases and agriculturally important traits is controlled by many genes with small effects. Identifying these genes, while simultaneously controlling false positives, is challenging. Among available statistical methods, the mixed linear model (MLM) has been the most flexible and powerful for controlling population structure and individual unequal relatedness (kinship), the two common causes of spurious associations. The introduction of the compressed MLM (CMLM) method provided additional opportunities for optimization by adding two new model parameters: grouping algorithms and number of groups.ResultsThis study introduces another model parameter to develop an enriched CMLM (ECMLM). The parameter involves algorithms to define kinship between groups (that is, kinship algorithms). The ECMLM calculates kinship using several different algorithms and then chooses the best combination between kinship algorithms and grouping algorithms.Conclusion Simulations show that the ECMLM increases statistical power. In some cases, the magnitude of power gained by using ECMLM instead of CMLM is larger than the improvement found by using CMLM instead of MLM.BMC biology. 10/2014; 12(1):73.
Dataset: 20103409414[Show abstract] [Hide abstract]
ABSTRACT: This paper describes the results of Chiari osteotomy (CO) with simultaneous intertrochanteric osteotomy (IO) in dogs affected by bilateral hip dysplasia, score grade D. The dogs classified for the study were subjected to the following tests before and after surgery: abduction-external rotation test, hip extension test, dorso-lateral subluxation test (DLST), stand test, Ortolani sign, Barlow sign, measurements of the angle of flexion, angle of extension, angle of abduction, angle of adduction, and radiographic examinations determining the score grade for canine hip dysplasia in accordance with the scoring system recommended by the Fédération Cynologique Internationale. The test which most accurately reflected the post-operative improvement in patients was DLST where the negative response increased by 73% after CO and IO procedures. Changes in angle of inclination (AI) values were correlated with an improvement in hip functioning as a result of the administered treatment. The simultaneous application of CO and IO in dogs affected by hip dysplasia resulted in greater improvement of limb functioning in comparison with conventional IO.
AJVR, Vol 70, No. 4, April 2009 483
osteoarthritis, lameness, and physical disability.1 Medi-
cal and surgical management of the condition have
economic and emotional impacts on dog owners and
breeders. As a complex trait, HD is caused by genetic
and environmental factors that influence expression of
the primary trait and the severity of secondary osteo-
arthritis.2 Factors that affect expression of HD and de-
velopment of secondary osteoarthritis in dogs include
sex, age, and body weight.2–6 Many genes likely under-
lie expression of HD, most of which have a small addi-
tive effect (polygenes), but some of which likely have
ip dysplasia in dogs is a polygenic disease charac-
terized by hip instability that results in secondary
Estimation of heritabilities, genetic correlations,
and breeding values of four traits
that collectively define hip dysplasia in dogs
Zhiwu Zhang, PhD; Lan Zhu, PhD; Jody Sandler, DVM; Steven S. Friedenberg, BS, MBA;
Jill Egelhoff, BS; Alma J. Williams, BS; Nathan L. Dykes, VMD; William Hornbuckle, DVM;
Ursula Krotscheck, DVM; N. Sydney Moise, DVM, MS; George Lust, PhD; Rory J. Todhunter, BVSc, PhD
Objective—To estimate heritabilities and genetic correlations among 4 traits of hip joints
(distraction index [DI], dorsolateral subluxation [DLS] score, Norberg angle [NA], and ex-
tended–hip joint radiograph [EHR] score) and to derive the breeding values for these traits
Animals—2,716 dogs of 17 breeds (1,551 dogs in which at least 1 hip joint trait was mea-
Procedures—The NA was measured, and an EHR score was assigned. Hip joint radio-
graphs were obtained from some dogs to allow calculation of the DI and DLS score. Heri-
tabilities, genetic correlations, and breeding values among the DI, DLS score, NA, and EHR
score were calculated by use of a set of multiple-trait, derivative-free, restricted maximum
likelihood computer programs.
Results—Among 2,716 dogs, 1,411 (52%) had an estimated inbreeding coefficient of 0%;
the remaining dogs had a mean inbreeding coefficient of 6.21%. Estimated heritabilities
were 0.61, 0.54, 0.73, and 0.76 for the DI, DLS score, NA, and EHR score, respectively. The
EHR score was highly genetically correlated with the NA (r = –0.89) and was moderately ge-
netically correlated with the DI (r = 0.69) and DLS score (r = –0.70). The NA was moderately
genetically correlated with the DI (r = –0.69) and DLS score (r = 0.58). Genetic correlation
between the DI and DLS score was high (r = –0.91).
Conclusions and Clinical Relevance—Establishment of a selection index that makes use
of breeding values jointly estimated from the DI, DLS score, NA, and EHR score should
enhance breeding programs to reduce the incidence of hip dysplasia in dogs. (Am J Vet Res
In North America, selection methods to improve the
genetic composition of dog breeds and, in so doing, im-
prove hip joint conformation have been based on radio-
graphic hip joint screening, semiopen (OFA) and closed
(PennHIP) hip joint registries, and organized breeding
programs.9–11 Through these breeding strategies, HD
was determined to be heritable, and selective breeding
efforts reduced the prevalence of HD.12 For example,
the prevalence of HD in German Shepherd Dogs at 12
to 16 months of age decreased from 55% to 24% after
5 generations of selection,12 and the prevalence of HD
in Labrador Retrievers decreased from 30% to 10%.13,14
Selective breeding is less effective when a single phe-
Received March 27, 2008.
Accepted July 22, 2008.
From the Institute for Genomic Diversity (Zhang), the Department
of Clinical Sciences (Friedenberg, Egelhoff, Williams, Dykes, Horn-
buckle, Krotscheck, Moise, Todhunter), and the Baker Institute for
Animal Health (Lust), College of Veterinary Medicine, Cornell Uni-
versity, Ithaca, NY 14850; the Department of Statistics, College of
Arts and Sciences, Oklahoma State University, Stillwater, OK 74078
(Zhu); and Guiding Eyes for the Blind, 611 Granite Springs Rd,
Yorktown Heights, NY 10598 (Sandler).
Address correspondence to Dr. Todhunter.
484 AJVR, Vol 70, No. 4, April 2009
notype is used as the selection criterion15–22 than when
estimated breeding values are used.23,24
Because HD has been so difficult to accurately de-
fine and eliminate in dogs, much effort has been di-
rected at developing and comparing the accuracy of
radiographic screening tests.3,25 The most widely used
method for diagnosis in North America is the ventro-
dorsal EHR, commonly referred to as the OFA meth-
od.26 From this radiographic image, a subjective EHR
score is obtained. Objective measurements include the
DI,27 DLS score,3,28,29 and NA.30 Heritabilities of these
traits reportedly range from 0.10 to 0.68.31
Because the objective hip joint traits and the EHR
score are modestly correlated with each other at the
phenotypic level,25,29,32,33 the estimates of heritabili-
ties and breeding values derived from a multiple-trait
model, which incorporates genetic and environmental
correlations among the traits, would be more accurate
than if they were derived from a single trait.34,35 More
importantly, the genetic correlations estimated from a
multiple-trait model would provide the essential values
by which a selection index could be derived to integrate
the breeding values of all the traits.36 Selective breed-
ing based on these combined breeding values should
be more effective in reducing the prevalence of such a
complicated trait than breeding decisions made on the
basis of a breeding value for a single trait.37
Although heritabilities of radiographic hip joint
measurements have been investigated in various dog
breeds in various environments,33,38–41 estimates of ge-
netic correlations among values used to judge hip joint
quality are limited. The purpose of the study reported
here was to estimate genetic correlations among the DI,
DLS score, NA, and EHR score and their heritabilities in
a multiple-trait model for subsequent use in deriving a
breeding value for each of these 4 traits of hip joints.
Materials and Methods
Animals—Dogs used in the study originated from
closed breeding colonies at the Baker Institute for Ani-
mal Health at Cornell University, the Guiding Eyes for
the Blind in Yorktown Heights, NY, or those admitted to
the Cornell University Hospital for Animals for radio-
graphic evaluation from January 1999 through October
2007. Multiple radiographic evaluations were available
for some dogs. The hip joints of the Baker Institute dogs
were commonly radiographed at 8 to 12 months of age.
The hip joints of the dogs at the Guiding Eyes for the
Blind were routinely radiographed at 14 to 18 months of
age. Dogs admitted to the Cornell University Hospital for
Animals were radiographed at any age > 8 months.
Hip dysplasia scoring—As is typical in North
America, an EHR was used to assess conformation of
each hip joint by assigning a subjective rating of ex-
cellent, good, or fair to the joint and borderline and
mild, moderate, or severe to the degree of hip dysplasia
evident (EHR score). The NA30 was measured from the
EHR and ranged from 50° (a subluxated hip joint) to
123° (a hip joint phenotypically unaffected by HD). The
maximum amount of lateral femoral head distraction
from the acetabulum (ie, the DI) was measured through
the PennHIP by means of a radiograph obtained with
the hip joint in the distraction position. Labrador Re-
trievers with a DI < 0.3 to 0.4 at 8 months of age were
presumed to have a > 80% probability of not develop-
ing secondary osteoarthritis in hip joints and were clas-
sified as unaffected by HD. Those with a DI > 0.7 were
presumed to have a high probability of developing os-
teoarthritis in hip joints and were classified as affected
with HD.42–46 The PennHIP also involves assessment of
EHRs to determine hip joint conformation and whether
secondary osteoarthritis exists. When no indication of
trauma was evident on the EHR, detection of secondary
osteoarthritis in hip joints was believed to be indica-
tive of antecedent HD. The DLS score was measured as
the percentage of femoral head covered by the dorsal
acetabulum with the hip joint in a natural, weight-bear-
Higher breeding values for the NA and DLS score
indicated a better hip joint (ie, less dysplastic), whereas
lower breeding values for the DI and EHR score in-
dicated the same thing. Body weight, breed, and sex
were recorded when the hip joints were radiographed.
Breeds represented by < 10 dogs were removed from
subsequent statistical analyses.
Pedigree—Ancestors of each dog were traced back
until no parent could be identified. Dogs for which hip
joint traits had not been measured were used to geneti-
cally connect the dogs from which measurements had
been obtained. The additive relationship (kinship) ma-
trix (2,716 X 2,716) was calculated from the pedigree
by means of the tabulate method.48,49 The calculation
began with dogs at the highest (earliest) generation
and carried all the way to the dogs without progeny.
The diagonals of the matrix were equal to 1 plus the
inbreeding coefficient.50 The inbreeding coefficient was
set at 0% for dogs with unknown parents or no com-
mon ancestor within the depth of pedigree tracked for
their parents. The kinship coefficient (ie, coancestry)
between each pair of dogs was also calculated and was
equivalent to the inbreeding coefficient of the hypo-
thetical progeny of each pair (ie, the probability that 2
alleles, sampled at random from each dog, were identi-
cal by descent).
Statistical analysis—Summary data regarding sig-
nalment of dogs are presented as mean ± SD. A mul-
tiple-trait model was used to fully explore relationships
among dogs and among hip joint traits. The most dys-
plastic hip joint of each dog (highest EHR score and
DI and lowest DLS score and NA of the 2 hip joints for
each dog) was used as the measurement. To improve
the accuracy for the prediction of breeding values for
each dog and its relatives, multiple measurements from
different ages were used whenever available. In addition
to providing estimates of genetic correlation among hip
joint traits, the multiple-trait model provided accurate
predictions of breeding values. In matrix notation, the
multiple-trait model was as follows:
y = Xβ + Zu + e,
in which y is the vector of phenotypic values for the
4 traits (DI, DLS score, NA, and EHR score), β is the
vector of fixed effects for sex and breed (categoric vari-
AJVR, Vol 70, No. 4, April 2009 485
ables) and age and body weight (continuous variables),
u is the vector of unknown random additive genetic ef-
fects (the estimate of u is referred to as the BLUP or the
breeding value), and e is the vector of residual terms.
The X and Z are known incidence matrices.
For the random effects, it was assumed that u was
normally distributed with a mean of 0 and a variance of
G, where G = G0 ⊗ A. It was also assumed that e was
normally distributed with a mean of 0 and a variance
of R, where R = R0 ⊗ I. The variables G0 and R0 are un-
known 4 X 4 genetic and residual covariance matrices,
respectively, for the 4 hip joint traits; A is the additive
relationship matrix; and I is the identity matrix. Opera-
tor ⊗ is the direct product of 2 matrices, which is also
referred to as the Kroneker or Zehfuss product. Con-
sequently, the covariance of y (V) is calculated as V =
ZGZT + R, where operator T is transposed. The estimate
of β and prediction of u are β = (XTV–1X) X (–XTV–1y) and
u = (GZTV–1) X (y –
and the symbol – represents general inverse.
Restricted maximum likelihood estimates were ob-
tained for unknown variables G0 and R0 by use of a set of
multiple-trait, derivative-free, restricted maximum like-
lihood software packages that contained 3 programs.51,a
Through use of the first program, the additive relation-
ship matrix was calculated directly from the pedigree;
then its inverse matrix and the determinant of the origi-
nal matrix were evaluated to estimate the log-likelihood
function. The inbreeding coefficient, defined as the
probability that 2 alleles at any locus are identical by
descent,52 was calculated for each dog. The inbreeding
coefficient was calculated by means of a tabular method
described elsewhere.48,49 The second program was used
to prepare coefficients for the mixed model equation
on the basis of a statistical model with fixed and ran-
dom factors for single- or multiple-trait analysis. The
third program was used to solve the mixed-factor linear
xβ), where the operator –1 is inverse
equation and calculated the estimates of the variance
components that maximized the restricted likelihood
given the phenotypic data.54
A single trait-by-trait analysis was conducted
first. The estimates of additive genetic variance and
residual variance were used as the starting values for
the 2-trait analysis on all pairwise combinations of
the DI, DLS score, NA, and EHR score. Variances from
the single-trait analyses and covariances estimated
from 2-trait analyses were used as the starting values
for the 4-trait analysis. Heritability was defined as
the ratio of the additive genetic variance to the total
variance (the sum of additive genetic variance and
residual variance). In the final multiple-trait analy-
sis, iterations were assumed to have converged when
the variance of –2 times the log-likelihood used in
the simplex search algorithm was < 10–9. To ensure
a global maximization in the log-likelihood, a re-
start of the computer programs was performed, with
the converged values used as the restarting points.
Restarts were performed until the difference of the
–2(log-likelihood) from 2 consecutive runs was
< 0.01. Estimates reported in the results section are
all from the 4-trait analysis. Breeding value accuracy
was estimated as the square root of (1 – PEV/σa
where PEV is the error variance of predicted breeding
values and σa
2 is the additive genetic variance.
Animals—The final data set contained 1,551
dogs with at least 1 of the 4 radiographic hip joint
measurements. Seventeen breeds were represented,
including Labrador Retriever, Greyhound, their cross-
breed offspring, and 14 others. Mean ± SD age of dogs
was 22.98 ± 22.11 months (range, 3 to 136 months).
Mean body weight of dogs was 29.30 ± 6.43 kg
(F1 X L) X (F1 X L)?
F1 X Greyhound?
F1 X L?
German Shepherd Dog? 4?
G = Generation depth. Nt = Total number of dogs. Nb = Number of dogs for which both parents were
known. Np = Number of dogs for which only 1 parent was known. Nu = Number of dogs for which neither par-
ent was known. Ni = Number of inbred dogs. Min = Minimum inbreeding value. Max = Maximum inbreeding
value. F1 = First filial generation resulting from a cross between a Labrador Retriever and a Greyhound. L =
Labrador Retriever. NA = Not available.
Table 1—Pedigree structure and inbreeding coefficient values within breed for dogs with a history of
HD from closed breeding colonies and a veterinary teaching hospital.
486 AJVR, Vol 70, No. 4, April 2009
(range, 5.7 to 57.7 kg). Males (46%) and females
(54%) were approximately equally represented.
Hip dysplasia scores—Most of the dogs were Lab-
rador Retrievers or their offspring from crossbreed-
ing with Greyhounds (F1, F1 backcrosses to Labrador
Retrievers or Greyhounds, and F2 offspring; Table 1).
Some traits were measured at multiple ages. The mean
number of measurements per dog was 1.3, 1.0, 1.2, and
1.1 for the DI, DLS score, NA, and EHR score, respec-
tively. The DLS score ranged from 85% for tight-hipped
Greyhounds to as low as 19% for the most dysplastic
dogs. The NA and EHR score were recorded for most
dogs, but the DI and DLS score were only available
in records from the Baker Institute for Animal Health
at Cornell University. Because the NA and EHR score
were measured on more dogs than were the DI and
DLS score, higher accuracy was expected for the NA
and EHR score. Representation of each variable among
the various breeds was summarized (Table 2). Values
for NA, DLS score, DI, and EHR score were also sum-
marized (Table 3). All 4 hip joint radiographic traits were
measured only for Labrador Retrievers, Greyhounds, and
their crossbreed offspring and German Shepherd Dogs.
For other breeds, only 1 to 3 traits were measured.
Pedigree—A total of 1,165 ancestors was added
to the pedigree, which contained 2,716 dogs, including
1,498 dogs from the Guiding Eyes for the Blind organiza-
tion, 571 from the Baker Institute for Animal Health, 425
from the Cornell University Hospital for Animals, and 222
ancestors traced from the database (Table 1). The role of
the 1,165 dogs without a measurement of hip joint quality
was to genetically connect the 1,551 dogs with measure-
ments. The most complex generation involved a family of
Labrador Retrievers from the Guiding Eyes for the Blind,
which included 1,236 connected dogs over 17 generations
from a particular male dog.
Among 2,716 dogs, about half (859 progeny and
552 founders; 53%) had an inbreeding coefficient of
0%. The remainder had a mean inbreeding coefficient
of 6.21%. The highest inbreeding coefficients (31.3%
and 37.7%) were obtained for only a few dogs. Inbreed-
ing coefficients and pairwise kinship coefficients were
summarized (Figures 1 and 2). Because mating did not
occur for any breeds other than Labrador Retriever and
Greyhound, > 80% of the total pairs of dogs had un-
known coancestries of 0.
Hip joint traits—The estimated effects of age,
sex, body weight, and breed on the 4 hip joint traits
were summarized (Table 4). Breed of dog had the
largest influence on the hip joint trait statistics. The
influence of dog age, body weight, and sex on the hip
joint traits was minimal. The mean NA for Labrador
Retrievers was not significantly different from that of
Australian Shepherds, Border Collies, Border Terri-
ers, Bullmastiffs, or Rottweilers. Whereas American
English Coonhounds, German Shepherd Dogs, Gold-
en Retrievers, Great Danes, Newfoundlands, and
Bullmastiffs had smaller NAs (ie, lower-quality hip
joints) than Labrador Retrievers (P < 0.01 for all),
Greyhounds had larger (P < 0.01) NAs (ie, higher-
(F1 X L) X (F1 X L)?
American English Coonhound?
F1 X Greyhound?
F1 X Labrador Retriever?
German Shepherd Dog?
Values in parentheses are the number of dogs evaluated in each category. Some dogs had 1 measure-
ment/type of score.
See Table 1 for remainder of key.
Table 2—Number of measurements used in the calculation of various scores for degree of HD in dogs
from closed breeding colonies and a veterinary teaching hospital that were radiographically evaluated
for hip dysplasia.
DLS score (%)
CV = Coefficient of variation.
See Table 1 for remainder of key.
Table 3—Values of the DI, DLS score, NA, and EHR score in dogs
from closed breeding colonies and a veterinary teaching hospital
that were radiographically evaluated for HD.
AJVR, Vol 70, No. 4, April 2009 487
quality hip joints). A similar pattern was evident for
The 4 hip joint traits had medium to high heritability
(Table 5). The EHR score had a high genetic correlation
with the NA (–0.89), and the genetic correlation between
the DI and DLS score was also high (–0.91). The medium
to high heritability for the 4 traits was also evident in the
pattern of breeding values during the years when dogs
were born (Figure 3). Selective breeding appeared effec-
tive at improving mean scores for the 4 hip joint traits over
time, particularly since the mid-1990s.
The accuracy of a breeding value for a dog was in-
fluenced by whether measurements were made on that
dog, whether the parents were measured, and the num-
ber of measured progeny (Table 6). The more progeny
of a breeding pair that were measured, the higher the
accuracy of the breeding value.
The results of selective breeding were also evident
in the relationship between breeding values and their
accuracy (Figure 4). Over half of the Labrador Retriev-
ers were bred at the Guiding Eyes for the Blind facil-
ity. Dogs with more accurate breeding values produced
more progeny, with a clustering of breeding values with
higher accuracy indicative of better hip joint conforma-
tion. In other words, such dogs had larger NAs and DLS
scores (positive breeding values) and smaller DIs and
EHR scores (negative breeding values), which indicat-
ed that the selective breeding practices of the Guiding
Eyes for the Blind program which are based on the NA
and EHR score, were effective in improving hip joint
conformation in dogs. Because the DI and DLS score
are genetically correlated to the NA and EHR score, an
indirect selection response on the DI and DLS score was
Figure 1—Distribution of inbreeding coefficients among 2,716
dogs from 2 closed breeding colonies and dogs admitted to a
veterinary teaching hospital.
Figure 2—Distribution of pairwise kinship coefficients among
2,716 dogs from 2 closed breeding colonies and dogs admitted
to a veterinary teaching hospital. The nonzero pairwise kinship
coefficient frequencies in the graph sum to 19.6%. The rest
(80.4%) were zeros and removed from the graph.
Table 4—Estimates of the fixed effects of age, body weight, sex, and breed on 4 traits of hip joints in
American English Coonhound?
F1 X Greyhound?
F1 X L?
German Shepherd Dog?
NC = Not calculated. – = Not applicable.
See Table 1 for remainder of key.
488 AJVR, Vol 70, No. 4, April 2009
also detected. The EHR score and NA were predicted
with highest accuracy, indicating more intensive selec-
tion was applied for these 2 traits.
The opportunity for reduction of the incidence of
HD in dogs by selective breeding on the basis of results
of hip joint radiographs has been available for over half
a century. In 1966, the OFA established a registry of
inherited orthopedic traits in dogs.55 Its initial mission
Table 5—Heritabilities (diagonals from top left to bottom right),
genetic correlations (values below diagonals), and residual cor-
relations (values above diagonals) among the DI, DLS score, NA,
and EHR score in 1,551 dogs from breeding colonies and a vet-
erinary teaching hospital that were evaluated for HD.
Percentage of dogs measured Mean No. of progeny
dogs Accuracy Self Sire Dam Per sire?
Table 6—Relationship between accuracy of breeding value and percentage of dogs that were mea-
sured on themselves, percentage of sires measured, percentage of dams measured, and mean num-
ber of progeny measured per sire and per dam in dogs evaluated for hip joint quality.
Figure 3—Mean breeding values for DI (A), DLS score (B), NA (C), and EHR score (D) estimated by means of the BLUP method
for 2,716 dogs (1,551 phenotyped dogs and 1,165 ancestors) from 2 closed breeding colonies and veterinary teaching hospital
born from 1966 through 2006. Lower or negative breeding values for the DI and EHR score reflect good hip joint conformation,
whereas higher (positive) breeding values for the DLS score and NA reflect the same thing.
AJVR, Vol 70, No. 4, April 2009 489
was to provide radiographic evaluation, data manage-
ment, and genetic counseling for reduction of the in-
cidence of HD. Since then, improvement in hip joint
phenotypes in North American purebred dogs has been
modest.15–19,22 This is partly attributable to the preferen-
tial submission of radiographs to the OFA for hip joint
certification, the preference of owners regarding publi-
cation of their dogs’ EHR scores on the OFA Web site,
and the difficulty breeders and dog owners have in mak-
ing use of all the pedigree and phenotype information
available in the OFA registry. Such semiopen databases
could be used to provide quality scores for hip joints
and breeding values of dogs that are based on genetic
correlations among several traits of hip joints. Howev-
er, one cannot ignore the fact that selective breeding on
the basis of hip joint phenotype alone would be more
effective if it were used by all breeders and purchasers
of purebred dogs. Lack of compliance by breeders and
owners with including hip joint phenotypes (ie, infor-
mation on hip joint traits) for some radiographed dogs
or not making such information available has led to
bias in the databases.22
The breeding value in its earliest use was also
called the selection index.56–58 The selection index was
based on integration of genetic and phenotypic infor-
mation from each animal and its relatives and yielded
better results than phenotypic selection alone for de-
sired traits. As was evident in the present study (Table
6), the accuracy of the selection index of a subject in-
creases when the phenotype information from its close
relatives (eg, progeny and ancestors) is included in the
estimation. With the assumption that environmental ef-
fects were perfectly estimated, which was not valid in
most situations, the selection index was developed into
the BLUP.59 The BLUP breeding strategy has been used
successfully for genetic improvement, particularly in
livestock, and has also been applied in closed colonies
Implementation of the BLUP strategy was enhanced
by methods of variance component estimation, such as
restricted maximum likelihood.60 Variance components
attributable to additive genetic and residual effects have
been estimated and the heritability of HD has been de-
rived in Finnish Rottweilers.40 Results of the present
study indicated that the establishment of a selection in-
dex that included the DI and DLS score along with the
NA and EHR score could be more effective in reducing
the incidence of HD than use of a single phenotypic
measurement. An EHR score and NA were measured in
most dogs, whereas the DI and DLS score were obtained
less commonly. Even so, we were able to infer the DI
and DLS score on the basis of the pedigree relationships
and genetic correlations among the hip joint traits. On
the other hand, most dogs with a DI and DLS score also
had an NA and EHR score, and this provided the es-
sential information to estimate the genetic correlations
The heritability of each hip joint trait estimated
in the present study was moderate to high for a com-
plex trait and may have been overestimated because of
the restricted population evaluated (many of the dogs
were reared in controlled environmental conditions).
Figure 4—Distribution of breeding values of 2,716 dogs from 2 closed breeding colonies and a veterinary teaching hospital compared
with the accuracy of those values for the DI (A), DLS score (B), NA (C), and EHR score (D). Lower or negative breeding values on the DI
and EHR score reflect better hip joint conformation as do higher (positive) breeding values for the DLS score and NA. Clustering of val-
ues at the top of the accuracy range (0.75 to 1.0) and clustering of values at the top of the accuracy range (0.75 to 1.0) and the right side
of breeding values for DLS scores and NAs or the left side for DIs and EHR scores is evidence of effective selection to reduce HD.
490 AJVR, Vol 70, No. 4, April 2009
Generally, most breeders and dog purchasers will only
have access to either an EHR score from the OFA or
a DI when they collaborate with a veterinarian who is
a member of the PennHIP. Although variable environ-
mental conditions can affect genetic improvement, se-
lection accuracy for genetic potential of hip joints with
good quality has been low in the general population of
dogs because most selective breeding is performed on
the basis of phenotype alone.
A preferable selection option for breeders and
purchasers of dogs would be selective breeding based
on genetic values of hip joint conformation or selec-
tion indices. The mean breeding values in our sample
of dogs during the study period clearly suggested the
genetic improvement that can be gained through selec-
tive breeding. In our study, the breeding value and its
accuracy for each dog were calculated for the DI, DLS
score, NA, and EHR score measured on the most dys-
plastic hip joint of 1,551 dogs. We identified a cluster
of dogs with highly accurate breeding values that also
indicated good hip joint conformation. The Labrador
Retrievers from the Guiding Eyes for the Blind program
constituted of over half of the dogs in the study. Se-
lection of dogs for hip joint quality resulted in genetic
improvement predominantly in the last 10 to 15 years,
when strong genetic selection pressure was applied on
dogs at the Guiding Eyes for the Blind on the basis of
phenotypes in the mid-1990s, and the BLUP method
was implemented to apply additional pressure on hip
joint conformation in 2004. Pedigrees with more ances-
tors and relatives also became available over the last 10
years, compared with the amount of data available in
the previous decade.
We propose selection indices derived from hip
joint traits of dogs in the public OFA database be ap-
plied when selecting of breeding dogs and puppies for
purchase. This approach should improve hip joint con-
formation of dogs in subsequent generations. Although
selective breeding based on choice of sires would be
more efficient than that based on choice of dams, selec-
tive breeding based on choice of both sexes would be
preferable. Breeding values and inbreeding coefficients
for the most recent generations of dogs in the OFA da-
tabase could be estimated and provided to the public
upon request so that dogs with optimal hip joint con-
formation could be selected as breeding dogs and for
Results of the multiple-trait modeling in the pres-
ent study strongly suggested that a single hip joint ra-
diograph does not provide as much information about
a dog’s genetic potential as a combination of measure-
ments of hip joint conformation. Thus, a single hip
joint measurement is insufficient to provide a basis for
breeding decisions. In the PennHip, a distraction pro-
jection is used to calculate the DI, and the EHR score is
used to assess whether HD and secondary osteoarthritis
exist. However, that database is not open. In another
study,33 we found that a combined NA and DLS score
or DI provided a more accurate prediction of secondary
osteoarthritis in hip joints than a single trait predictor
in young adult Labrador Retrievers, Greyhounds, and
their crossbreed offspring.33 If the DI and EHR scores
accumulated for each dog in the PennHIP were com-
bined into a selection index and made available in a
user-friendly interface for dog owners and breeders, ad-
ditional improvement in hip joint conformation would
likely ensue. As a start, such information could be pro-
vided to dog owners as part of the information they re-
ceive when their dog’s hip joints are evaluated.
As genetic markers become available for quan-
titative trait locus detection, genetic selection based
on results of BLUP estimated from phenotypes could
be developed into marker-assisted selection based on
results of BLUP estimated from marker genotypes.
Marker-assisted selection is superior to genetic selec-
tion by means of BLUP based on phenotypes alone.61,62
The BLUP based on genetic markers of susceptibility
and resistance to HD will be identified in the future,
and when identified, this information could be com-
bined with phenotype breeding values to improve se-
lective breeding programs. The high genetic correlation
between the EHR score and NA and between the DLS
score and DI in our study suggested that common genes
may be associated with these 2 pairs of hip joint traits.
In support of this supposition is the preliminary finding
from a molecular genetics study7 that some quantitative
trait loci identified as contributing to HD for a single
hip joint trait have pleiotropic effects on other hip joint
traits or that single traits are associated with linked loci.
This suggests that identifying genes that underlie the
NA and EHR score and the DI and DLS score will lead
to the most complete understanding of the molecular
genetic basis of HD. The addition of genetic marker
breeding values for the DI or DLS score to a hip joint
selection index should additionally improve hip joint
quality because some of the underlying mutations that
contribute to these 2 traits are likely different from
those underlying the NA and EHR score.
In the present study, proportions of records that
were missing for each of the 4 hip joint traits varied.
Only 11% of the 1,551 dogs in which at least 1 value
was measured were not measured for NA (4%) and EHR
score. Most dogs, particularly those born before 1990
or those from the Guiding Eyes for the Blind program,
were not measured for the DI (73%) and DLS score
(75%). Missing values were allowed in the multiple-
trait model. A dog with a missing value for 1 hip joint
trait was still assigned a predicted value for that trait
on the basis of other available measurements and the
same trait measured in its relatives. Therefore, a breed-
ing value for all 4 hip joint traits could be estimated for
a dog with no phenotype. Consequently, the breeding
values of the DI and DLS score of any given dog could
be inferred by the genetic correlation between the NAs
and EHR scores of all its relatives without these 2 traits
being measured. This is an important consideration
given that hip joint radiographs are not available for
some dogs in a pedigree from the semiopen OFA data-
base, but their breeding value for a hip joint trait can
still be estimated.
Thorough documentation of the pedigree would
also improve the estimation of inbreeding coefficients
and breeding values. Among 1,411 noninbred dogs in
the present study, there were 552 dogs for which no par-
ents were identified and 342 dogs with only 1 known
parent. Nevertheless, because breeding values can be
AJVR, Vol 70, No. 4, April 2009 491
estimated for dogs when no hip joint trait information
is available, users of semiopen registries can still benefit
from active participation because pedigree relationships
allow breeding values for hip joints of nonparticipating
dogs to be estimated.
When selecting dogs on the basis of breeding values
for hip joint conformation, some consideration should
be given to their inbreeding coefficients so genetic di-
versity is not adversely affected. Use of breeding val-
ues for hip joint conformation could lead to intensive
use of dogs with good breeding values, thus narrowing
genetic diversity. The effects of inbreeding would then
accumulate at a higher rate. An efficient mating system
would make use of predictions of the inbreeding co-
efficient that would result for progeny from mating a
particular pair. Inbreeding coefficients for a potential
breeding pair could be provided upon request simulta-
neously with breeding values to breeders and owners. A
breeder may consider all available breeding values and
calculate a selection index based on traits such as be-
havior, body size, and many orthopedic characteristics
as long as they were measured and recorded. Each trait
could be weighted for calculating the selection index
on the basis of its importance to the breed organiza-
tion. Given that mate pairing is also a factor in selec-
tive breeding, a male considered strong with respect to
a trait could be paired with a female that is weak with
respect to the same trait.
a. MTDFREML, Agricultural Research Service, USDA, Washing-
ton, DC. Available at: aipl.arsusda.gov/curtvt/mtdfreml.html.
Accessed Dec 29, 2008.
1. Todhunter RJ, Lust G. Canine hip dysplasia: pathogenesis. In:
Slatter D, ed. Textbook of small animal surgery. 3rd ed. Philadel-
phia: WB Saunders Co, 2003;2009–2019.
Kealy RD, Lawler DF , Ballam JM, et al. Evaluation of the effect
of limited food consumption on radiographic evidence of osteo-
arthritis in dogs. J Am Vet Med Assoc 2000;217:1678–1680.
Lust G, Todhunter RJ, Erb HN, et al. Repeatability of dorsolateral
subluxation scores in dogs and correlation with macroscopic ap-
pearance of hip osteoarthritis. Am J Vet Res 2001;62:1711–1715.
Chase K, Lawler DF , Adler FR, et al. Bilaterally asymmetric ef-
fects of quantitative trait loci (QTLs): QTLs that affect laxity in
the right versus left coxofemoral (hip) joints of the dog (Canis
familiaris). Am J Med Genet A 2004;124A:239–247.
Chase K, Lawler DF , Carrier DR, et al. Genetic regulation of os-
teoarthritis: a QTL regulating cranial and caudal acetabular os-
teophyte formation in the hip joint of the dog (Canis familiaris).
Am J Med Genet A 2005;135:334–335.
Vanden Berg-Foels WS, Todhunter RJ, Schwager SJ, et al. Ef-
fect of early postnatal body weight on femoral head ossification
onset and hip osteoarthritis in a canine model of developmental
dysplasia of the hip. Pediatr Res 2006;60:549–554.
Todhunter RJ, Mateescu R, Lust G, et al. Quantitative trait loci
for hip dysplasia in a cross-breed canine pedigree. Mamm Ge-
Zhu L, Zhang Z, Schweitzer P, et al. Single nucleotide polymor-
phisms refine QTL intervals for hip joint laxity in dogs. Anim
Ohlerth S, Busato A, Gaillard C, et al. Epidemiologic and ge-
netic studies of canine hip dysplasia in a population of Labra-
dor retrievers: a study over 25 years. Dtsch Tierarztl Wochenschr
10. Wood JL, Lakhani KH, Dennis R. Heritability and epidemiology of
canine hip-dysplasia score in flat-coated retrievers and Newfound-
lands in the United Kingdom. Prev Vet Med 2000;46:75–86.
11. Wood JL, Lakhani KH, Dennis R. Heritability of canine hip-dys-
plasia score and its components in Gordon setters. Prev Vet Med
12. Swenson L, Audell L, Hedhammar A. Prevalence and inheri-
tance of and selection for hip dysplasia in seven breeds of dogs
in Sweden and benefit: cost analysis of a screening and control
program. J Am Vet Med Assoc 1997;210:207–214.
13. Leighton EA, Linn JM, Willham RL, et al. A genetic study of
canine hip dysplasia. Am J Vet Res 1977;38:241–244.
14. Leighton EA. Genetics of canine hip dysplasia. J Am Vet Med As-
15. Corley EA, Hogan PM. Trends in hip dysplasia control: analy-
sis of radiographs submitted to the Orthopedic Foundation for
Animals, 1974 to 1984. J Am Vet Med Assoc 1985;187:805–809.
16. Corley EA. Hip dysplasia: a report from the Orthopedic Founda-
tion for Animals. Semin Vet Med Surg (Small Anim) 1987;2:141–
17. Corley EA. Role of the Orthopedic Foundation for Animals in
the control of canine hip dysplasia. Vet Clin North Am Small
Anim Pract 1992;22:579–593.
18. Corley EA, Keller GG, Lattimer JC, et al. Reliability of early ra-
diographic evaluations for canine hip dysplasia obtained from
the standard ventrodorsal radiographic projection. J Am Vet Med
19. Kaneene JB, Mostosky UV, Padgett GA. Retrospective cohort
study of changes in hip joint phenotype of dogs in the United
States. J Am Vet Med Assoc 1997;211:1542–1544.
20. Willis MB. A review of the progress in canine hip dysplasia con-
trol in Britain. J Am Vet Med Assoc 1997;210:1480–1482.
21. Rettenmaier JL, Keller GG, Lattimer JC, et al. Prevalence of ca-
nine hip dysplasia in a veterinary teaching hospital population.
Vet Radiol Ultrasound 2002;43:313–318.
22. Paster ER, LaFond E, Biery DN, et al. Estimates of prevalence
of hip dysplasia in Golden Retrievers and Rottweilers and the
influence of bias on published prevalence figures. J Am Vet Med
23. Langeland M, Lingaas F . Spondylosis deformans in the boxer:
estimates of heritability. J Small Anim Pract 1995;36:166–169.
24. Leppanen M, Saloniemi H. Controlling canine hip dysplasia in
Finland. Prev Vet Med 1999;42:121–131.
25. Lust G, Todhunter RJ, Erb HN, et al. Comparison of three radio-
graphic methods for diagnosis of hip dysplasia in eight-month-
old dogs. J Am Vet Med Assoc 2001;219:1242–1246.
26. Tsai KL, Murphy KE. Clinical and genetic assessments of hip
joint laxity in the Boykin spaniel. Can J Vet Res 2006;70:148–
27. Smith GK, LaFond E, Gregor TP, et al. Within- and between-
examiner repeatability of distraction indices of the hip joints in
dogs. Am J Vet Res 1997;58:1076–1077.
28. Farese JP, Lust G, Williams AJ, et al. Comparison of measure-
ments of dorsolateral subluxation of the femoral head and max-
imal passive laxity for evaluation of the coxofemoral joint in
dogs. Am J Vet Res 1999;60:1571–1576.
29. Farese JP, Todhunter RJ, Lust G, et al. Dorsolateral subluxation
of hip joints in dogs measured in a weight-bearing position with
radiography and computed tomography. Vet Surg 1998;27:393–
30. Olsson SE. Roentgen examination of the hip joints of German
Shepherd Dogs. Adv Small Anim Pract 1961;3:117–118.
31. Breur GJ, Lust G, Todhunter RJ. Genetics of hip dysplasia and
other orthopedic traits. Chapter 9. In: Ruvinsky A, Sampson J,
eds. Genetics of the dog. Wallingford, England: CAB Internation-
32. Puerto DA, Smith GK, Gregor TP, et al. Relationships between
results of the Ortolani method of hip joint palpation and dis-
traction index, Norberg angle, and hip score in dogs. J Am Vet
Med Assoc 1999;214:497–501.
33. Todhunter RJ, Grohn YT, Bliss SP, et al. Evaluation of multiple
radiographic predictors of cartilage lesions in the hip joints of
eight-month-old dogs. Am J Vet Res 2003;64:1472–1478.
34. Jiang C, Zeng ZB. Multiple trait analysis of genetic mapping for
quantitative trait loci. Genetics 1995;140:1111–1127.
35. Sapp RL, Rekaya R, Misztal I, et al. Longitudinal multiple-trait
versus cumulative single-trait analysis of male and female fer-
492 AJVR, Vol 70, No. 4, April 2009
tility and hatchability in chickens. Poult Sci 2005;84:1010–
36. Ohlerth S, Lang J, Busato A, et al. Estimation of genetic popula-
tion variables for six radiographic criteria of hip dysplasia in a
colony of Labrador Retrievers. Am J Vet Res 2001;62:846–852.
37. Carnier P, Gallo L, Sturaro E, et al. Prevalence of spondylosis
deformans and estimates of genetic parameters for the degree
of osteophytes development in Italian Boxer dogs. J Anim Sci
38. Hedhammar A, Olsson SE, Andersson SA, et al. Canine hip dys-
plasia: study of heritability in 401 litters of German Shepherd
dogs. J Am Vet Med Assoc 1979;174:1012–1016.
39. Mackenzie SA, Oltenacu EA, Leighton E. Heritability estimate for
temperament scores in German shepherd dogs and its genetic cor-
relation with hip dysplasia. Behav Genet 1985;15:475–482.
40. Maki K, Liinamo AE, Ojala M. Estimates of genetic parameters
for hip and elbow dysplasia in Finnish Rottweilers. J Anim Sci
41. Janutta V, Hamann H, Klein S, et al. Genetic evaluation of elbow
angles as predictors of elbow dysplasia in German shepherd
dogs. J Vet Med A Physiol Pathol Clin Med 2005;52:254–261.
42. Lust G, Williams AJ, Burton-Wurster N, et al. Joint laxity and its
association with hip dysplasia in Labrador retrievers. Am J Vet
43. Popovitch CA, Smith GK, Gregor TP, et al. Comparison of sus-
ceptibility for hip dysplasia between Rottweilers and German
Shepherd Dogs. J Am Vet Med Assoc 1995;206:648–650.
44. Smith GK, Popovitch CA, Gregor TP, et al. Evaluation of risk
factors for degenerative joint disease associated with hip dyspla-
sia in dogs. J Am Vet Med Assoc 1995;206:642–647.
45. Smith GK, Laster LL, Biery DN, et al. OFA reliability figures
questioned. J Am Vet Med Assoc 1998;212:487–488.
46. Smith GK, Mayhew PD, Kapatkin AS, et al. Evaluation of risk fac-
tors for degenerative joint disease associated with hip dysplasia in
German Shepherd Dogs, Golden Retrievers, Labrador Retrievers,
and Rottweilers. J Am Vet Med Assoc 2001;219:1719–1724.
47. Burton-Wurster N, Farese JP, Todhunter RJ, et al. Site-spe-
cific variation in femoral head cartilage composition in dogs
at high and low risk for development of osteoarthritis: in-
sights into cartilage degeneration. Osteoarthritis Cartilage
48. Cruden D. The computation of inbreeding coefficients for
closed populations. J Hered 1949;40:248–251.
49. Emik LO, Terrill CE. Systematic procedures for calculating in-
breeding coefficients. J Hered 1949;40:51–55.
50. Falconer DS, Mackay TFC. Introduction to quantitative genetics.
London: Prentice Hall, 1996.
51. Boldman KG, Kriese LA, Van Vleck LD, et al. A manual for use
of MTDFREML. A set of programs to obtain estimates of variances
and covariances. Washington, DC: Agricultural Research Ser-
vice, USDA, 1993.
52. Wright S. Coefficients of inbreeding and relationships. Am Nat
53. Tier B. Computing inbreeding coefficients quickly. Genet Sel
54. Zhang Z, Todhunter RJ, Buckler ES, et al. Technical note: use
of marker-based relationships with multiple-trait derivative-free
restricted maximal likelihood. J Anim Sci 2007;85:881–885.
55. Orthopedic Foundation for Animals. Search OFA records [data-
base online]. Available at: www.offa.org/search.html. Accessed
Dec 12, 2008.
56. Lush JL. Genetics and animal breeding. J Hered 1936;27:201–203.
57. Hazel LN, Lush JL. The efficiency of three methods of selection.
J Hered 1942;33:393–399.
58. Lush JL. The impact of genetics on animal breeding. J Anim Sci
59. Van Vleck LD. C. R. Henderson: farm boy, athlete, and scientist.
J Dairy Sci 1991;74:4082–4096.
60. Patterson HD, Thompson R. Recovery of inter-block information
when block sizes are unequal. Biometrika 1971;58:545–554.
61. Zhang W, Smith C. Computer simulation of marker-assisted
selection utilizing linkage disequilibrium. Theor Appl Genet
62. Zhang W, Smith C. Simulation of marker-assisted selection uti-
lizing linkage disequilibrium: the effects of several additional
factors. Theor Appl Genet 1993;86:492–496.