Genetic Analysis of Ventricular Arrhythmia in Young German
Shepherd Dogs
J. Cruickshank, R.L. Quaas, J. Li, S. Hemsley, T.M. Gunn, and N.S. Moı¨se
Background: Ventricular arrhythmias (VA) and sudden death are inherited in German Shepherd Dogs (GSDs).
Objectives: To estimate the genetic parameters (heritabilities and correlations) of 3 traits of VA (single premature ventric-
ular complexes (PVCs), 2 consecutive PVCs (couplets), and 3 or more consecutive PVCs—ventricular tachycardia [VT]).
Animals: Three hundred and ninety-eight GSDs.
Methods: Prospective, observational, case control study. Dogs were phenotyped by 24-hour ambulatory ECG from 6 to 45
weeks of age. Edited ECG records included the number of incidents of (1) single PVCs, (2) couplets, and (3) VT.
Results: A data set of 1,239 Holter records from 398 GSDs was used to estimate genetic variables. Phenotypic correlations
for affectedness (binarily coded 0/1) of the 3 traits ranged from 0.55 to 0.74, whereas correlations for severity (continuous values
of 24-hour VA counts) ranged from 0.26 to 0.39. Estimates of genetic correlation among the severity traits were 0.06 to 0.27.
Estimated heritabilities were 0.54, 0.54, and 0.46 for affectedness and 0.33, 0.69, and 0.69 for severity of PVCs, couplets, and
VT, respectively. Month and year of birth and age at ECG recording had significant effects on all 3 traits. Season of ECG
recording had a significant effect on the number of single PVCs, but not couplets or incidents of VT. Age of onset differed, with
single PVCs appearing an average of 4 days earlier than couplets and VT.
Conclusion: These results imply a strong genetic component for this disease but suggest that differences in the 3 traits should
be taken into consideration in studies to identify the underlying genes.
Key words: Canine; ECG in dogs; Holter monitoring; Inherited heart disease; Model of arrhythmias.
S
pontaneous ventricular arrhythmia (VA) and sudden
death occur in young German Shepherd Dogs
(GSDs).
1–5
The disorder ranges in severity from infrequent
and non-life-threatening single premature ventricular com-
plexes (PVCs) to multiple episodes of rapid polymorphic
ventricular tachycardia (VT). Dogs with VT are most likely
to die suddenly.
1,6,7
No other clinical indicators of abnor-
mality are apparent, and pathological examination of the
hearts of dogs that die suddenly reveal morphologically
normal hearts.
1
A window of vulnerability for the presence
of VA and sudden death exists between approximately 3
and 18 months of age, with peak affectedness occurring at
approximately 6–7 months of age.
1,6
Affected dogs rarely
have VA after 24 months of age, and when they do, it is
infrequent.
1,6
The trait of VT is most commonly observed
in dogs lying at rest and during rapid eye movement sleep.
8
Because of the age and behavioral dependence of the
expression of this disease, extensive observation via
24-hour ambulatory ECG monitoring (Holter monitoring)
is often required to ascertain disease presence and its
severity.
In people, important advances have been made in the
understanding of the genetic basis of inherited arrhythmic
disorders with loss of repolarization reserve including
the long QT syndromes,
9–12
Brugada syndrome,
13
and
others.
14–19
Causative mutations have primarily been
identified in genes encoding specific sodium and potas-
sium channel proteins involved in repolarization.
9–19
In
affected GSDs, abnormalities of ion current densities of
potassium channels and calcium cycling indicate a global
disruption of repolarization reserve.
20–30
Futhermore, het-
erogeneous sympathetic innervation has been identified.
29
Given the failure to identify a singular abnormality, ge-
nome-wide analysis of linkage is more likely to be
successful in identifying the causative loci for VA in GSDs
than a candidate gene approach. In anticipation of con-
ducting a full genome scan for loci linked to the disease,
we performed a genetic characterization of this trait in our
closed colony of GSDs with a high incidence of VA. Our
goals for this study were to estimate the genetic parame-
ters of VA traits and to characterize the progression of the
arrhythmia in this population.
Materials and Methods
Dogs
Dogs were housed in an American Association for the Accredi-
tation of Laboratory Animal Care approved facility, which
conforms to the guidelines for proper animal care as described
within the National Institutes of Health Guide for the Care and Use
of Laboratory Animals. Experiments were approved by the Institu-
tional Animal Care and Use Committee of Cornell University. Data
were collected on GSDs within a long-established (18 years) colony
at Cornell University’s College of Veterinary Medicine. Primary
founders of this colony came from families with incidences of sud-
den death that, under subsequent analysis, were shown to have a
common ancestor.
1
Additional affected and unaffected dogs have
been brought into the colony over the years, in part to diversify the
breeding pool. Thirty-eight backcross progeny (GSD  [GSD 
greyhound]) were used for pedigree analysis, but only data from
purebred GSDs were used for the quantitative analysis presented in
the current study. Dogs were born during 1988–2006, and pedigree
information was available for all dogs with phenotypic data. In-
From the Department of Biomedical Sciences (Cruickshank,
Gunn), the Department of Clinical Sciences (Hemsley, Moı¨se),
College of Veterinary Medicine, and the Department of Animal
Science (Quaas, Li), Cornell University, Ithaca, NY.
Corresponding author: N. Sydney Moı¨se, Department of Clinical
Sciences, College of Veterinary Medicine, Cornell University, Ithaca,
NY 14853; e-mail: nsm2@cornell.edu.
Submitted July 29, 2008; Revised October 29, 2008; Accepted
December 1, 2008.
Copyright
r
2009 by the American College of Veterinary Internal
Medicine
10.1111/j.1939-1676.2009.0265.x
J Vet Intern Med 2009;23:264–270

breeding coefficients were calculated for all dogs using ‘‘proc in-
breed’’ in SAS (v. 9.1, SAS Institute, Cary, NC).
Phenotype Analysis
Ambulatory ECG (Holter monitoring) was performed on all
dogs. Holter monitoring was serially performed with a median of
3.5 recordings each (eg, each dog evaluated at 11, 15, and 18 weeks
of age), although dogs born between 1987 and 1997 had more ex-
tensive monitoring in order to characterize the arrhythmogenic
phenotype. All ECG recordings were edited for correct beat anno-
tation by experienced technicians. Recordings from dogs under 14
weeks of age were 15 hours rather than 24 hours in duration and the
acquisition frequency was 1,000 Hz rather than 400 Hz. All records
were normalized to 24 hours to enable comparison of data among
all recordings. The term ‘‘record’’ refers to 1 dog’s ECG data re-
corded at 1 age. No treatment was administered before or after
Holter monitoring.
Recordings were made at ages of 3–413 weeks. Because of the
small number of ECG recordings on dogs under 6 and over 45
weeks of age, those records were removed from this analysis. The
frequency of VA in each ECG recording was identified and 3 VA
traits were categorized: single PVCs, 2 consecutive PVCs (couplets),
and 3 or more consecutive PVCs (VT). Some recordings analyzed
with earlier versions of software had such extensive VA that it was
not possible to annotate all beats. In order to give a semiquantifica-
tion of VA, thresholds for beat identification had to be made. In
these records, a threshold of 5,000 PVCs was set as a maximum;
such values were treated as missing data for analyses involving se-
verity. For the analysis of age of onset, data from dogs born in 1988
and 1989 were discarded because ECG recordings were not rou-
tinely collected before 18 weeks of age in those years.
Dogs were categorized as affected or unaffected for each of the 3
traits based on their most severe record for each trait. The cutoff
values for affectedness were 4100 PVCs, 44 couplets, and 42VT
in 24 hours. Although sudden death occurred in this population of
dogs, it was not included in the analysis because not all dogs were
kept beyond 6 months of age and were therefore not given the full
opportunity to express this trait.
Statistical Models
Descriptive statistics and Pearson’s correlation coefficients for
affectedness, severity, and age of onset of the VA traits (PVCs, cou-
plets, and VT) were calculated by SAS (v. 9.1, SAS Institute).
Significant fixed effects were identified with univariate analyses.
Variance and covariance components were estimated by restricted
maximum likelihood and procedures in SAS (v. 9.1, SAS Institute).
Heritability (h
2
) was estimated by
h
2
¼ s
2
a
=ðs
2
a
þ s
2
e
Þ;
where s
2
a
is the additive genetic variance and s
2
e
is the residual vari-
ance.
Animals often share common ancestors and consequently share
identity by descent, and animal effects are commonly fitted as ran-
dom effects. Unfortunately, these characteristics are not included in
any of the predefined variance structures by SAS (SAS Institute)
and are therefore ignored in empirical analyses. We developed an
SAS macro, MTCC, on the basis of LORG (coded by Zhiwu Zhang,
http://www.people.cornell.edu/pages/zz19/research/LORG/index.
html), to incorporate the genetic relationships among the dogs. This
macro computes numerical relationship among individuals from a
given pedigree using half storage, extracts the numerical relation-
ship, and assigns genetic parameters in sparse format to generate an
SAS data set that can be directly used by options GDATA or LDA-
TA for multiple trait data.
Affectedness (for PVCs, couplets, and VT coded as binary, un-
affected/affected) was analyzed with SAS (v. 9.1) ‘‘proc glimmix,’’
using the model
PrðY ¼ 1Þ¼yðmþ dogGþ eÞ;
where Pr(Y 5 1) denotes the probability a dog is affected; y is the
cumulative normal distribution function, which is the probit link
between the linear predictor on the right-hand side and the proba-
bility for a specific outcome on the left-hand side; m is a fixed model
intercept; dogG is the random genetic effect of dog, (0,As
2
G
)
where A is the additive genetic relationship matrix and s
2
G
is the ad-
ditive genetic variance; and e is the random residual error.
Preliminary analyses included terms for sex, birth year, and in-
breeding, but none of these was significant (P4 .10), and thus they
were not included in subsequent analyses.
Age of onset for the 3 severity traits was analyzed using ‘‘proc
mixed’’ in SAS (v. 9.1) using the following animal model:
Y ¼ mþ byearþ traitþ dogGþ dogPEþ e;
where Y is age of onset; m is a fixed model intercept; byear is the
dog’s year of birth; trait indicates as to which trait the age corre-
sponds to; dogG is the random additive genetic effect of dog, (0,
As
2
G
); dogPE is the random permanent effect of dog, (0, Is
2
PE
)
where I is an identity matrix and s
2
PE
is the permanent environmen-
tal variance; and e is the random residual error.
Severity traits (PVCs, couplets, or VT) were analyzed with SAS
(v. 9.1) ‘‘proc mixed,’’ using the model
logðYÞ¼mþ bmonyearþ seasonþðb
1
Þageþðb
2
Þage
2
þ dogG
þ dogPEþ e;
where Y is the number of PVCs, couplets, or VT in the ECG re-
cording; m is a fixed model intercept; bmonyear is the dog’s month
and year of birth; season is the season of ECG recording (4 seasons,
where January-March is season 1); fit as a covariate, age is the age in
weeks of dog at ECG recording; dogG is the random additive ge-
netic effect of dog, (0, As
2
G
); dogPE is the random permanent
effect of dog, (0, Is
2
PE
); and e is the random residual error. Pre-
liminary analyses included terms for sex and inbreeding, but neither
term was significant (Po .10) and therefore they were not included
in subsequent analyses. Covariance components between each 2
traits were obtained from bivariate analyses.
Results
Three hundred and ninety-eight GSDs with 1,239 re-
cords were included in the analysis. Inbreeding
coefficients for the dogs ranged from 0 to 0.375, with an
average of 0.126. The minimum VA count recorded for
all traits was 0. Across all records, the maximum re-
corded VA counts were 88,234 PVCs, 30,068 couplets,
and 14,205 incidents of VT. Other summary statistics are
presented in Table 1.
Table 1. Proportion of German Shepherd Dogs (n 5
398) affected and summary statistics of ventricular ar-
rhythmia data.
PVCs Couplets VT
% Affected 69 56 51
Range 0–88,234 0–30,068 0–14,205
Mean 2,373 223 48
Standard deviation 7,731 1,527 502
PVCs, single premature ventricular complexes; couplets, 2 con-
secutive PVCs; VT, ventricular tachycardia.
265Genetics of Arrhythmias in German Shepherds