INTEROBSERVER VARIABILITY OF VERTEBRAL HEART SIZE MEASUREMENTS IN DOGS WITH NORMAL AND ENLARGED HEARTS KERSTIN HANSSON, JENS HA¨GGSTRo¨M, CLARENCE KVART, PETER LORD
The vertebral heart size (VHS) method by Buchanan is based on anatomic landmarks. A potential source of variation among observers is differences in the selection of measurement points. The objective was to test variability between observers with different levels of training in thoracic radiology and small animal clinical practice. Fifty sets of thoracic radiographs of cavalier King Charles spaniels, were divided into five groups; (Normal) normal cardiopulmonary structures, (I) slight cardiomegaly, (II) moderate cardiomegaly, (II þ) moderate cardiomegaly with congestive heart failure, and (III þ) severe cardiomegaly with congestive heart failure. Cardiomegaly was confirmed by echocardiography to be caused by mitral regurgitation because of myxomatous mitral valve disease. Sixteen observers representing four levels of experience (four observers/level) evaluated the radiographs; (1) European Diplomates in Veterinary Diagnostic Imaging, (2) Experienced small animal clinicians, (3) Trainees in small animal clinical practice (4) Veterinary students. Almost identical mean VHS values were found among the four experience levels for each of the five groups of radiographs with a low coefficient of variation, range 1.5–3.2%. Mean difference among the 16 observers was 1.05 0.32 vertebrae (v). Mean difference among individuals in each observer group was approximately 0.5 v for all but the groups of trainees were the difference was 0.6–0.9 v. The conclusion is that VHS method for heart size is independent of observer experience but dependent of individual observers selection of reference points and transformation of long and short axis dimensions into VHS units. Veterinary Radiology & Ultrasound, Vol. 46, No. 2, 2005, pp 122–130.
Key words: dog, interobserver, mitral regurgitation, thoracic radiology, vertebral heart size.
Introduction
T
well-defined measurement points and objective numerical measurement. It uses the sum of the heart length and width, which is translated into total units of thoracic vertebral length (v) to the nearest 0.1 v. Based on evaluation of 100 clinically normal dogs of various breeds the mean VHS was 9.7 0.5 v. As the VHS was 10.5 in 98% of the dogs this was suggested as a clinically useful upper limit for normal heart size in most breeds.10 Lamb et al. has published breed-specific ranges for six breeds in which mean VHS was higher than 10.5 in three out of the six breeds; cavalier King Charles spaniel (10.6 0.5 v), Labrador retriever (10.8 0.6 v) and Boxer (11.6 0.8 v).11 If normal VHS values and ranges are to be generally applicable it is necessary that the interobserver variability be low among a typical group of users, which will include veterinarians inexperienced as well as experienced in thoracic radiology. Nakayama12 found a low coefficient of variation (CV) among observers, but the study included only observers who were experienced interpreters of thoracic radiographs. In a study by Lamb including three observers with different experience (veterinary radiologist, medicine resident, and veterinary nurse), the observers recorded similar mean values but a maximum likely difference of 41.0 v.13 As the VHS method is based on anatomic landmarks a potential source of variation between observers is that
commonly used in dogs to evaluate suspected cardiomegaly. Differences in age, conformation, cardiac cycle, and respiratory phase are causes of normal variation that make it difficult to separate normal from enlarged hearts.1–5 However, the inexperienced observer may have difficulties in identifying cardiomegaly even under the best of circumstances. The trained interpreter can often recognize cardiomegaly based on experience but inexperienced veterinarians may benefit from a quantitative method to evaluate heart size. Various methods have been used to estimate or measure heart size including planimetry and various cardio-thoracic ratios.3,6,7 Two frequently used methods for the lateral view are the intercostal8,9 and the vertebral heart size (VHS) method.10 The VHS method has the advantage of HORACIC RADIOLOGY IS
From the Departments of Biomedicine and Veterinary Public Health, Division of Diagnostic Imaging and Clinical Pathology (Hansson, Lord), Small Animal Medicine and Surgery (Ha¨ggstro¨m) and Animal Physiology (Kvart), Swedish University of Agricultural Sciences, Uppsala, Sweden. Address correspondence and reprint requests to: Kerstin Hansson, Department of Biomedicine and Veterinary Public Health, Division of Diagnostic Imaging and Clinical Pathology, Swedish University of Agricultural Sciences, Box 7029, SE-750 07 Uppsala, Sweden. E-mail:
[email protected] Received April 29, 2003; accepted for publication September 4, 2004. doi: 10.1111/j.1740-8261.2005.00024.x
122
Vol. 46, No. 2
123
InterobserverVariability, Heart Size by VHS
varying degrees of knowledge of thoracic radiographic anatomy might influence how the observer defines the various measurement points. Our objective was to study the variability of the VHS method in dogs with a range of different heart sizes among observers with different levels of experience in thoracic radiology.
Material and Methods Material The material consisted of 50 sets of thoracic radiographs of privately owned cavalier King Charles spaniels. All dogs were radiographed at the Department of Biomedicine and Veterinary Public Health, Division of Diagnostic Imaging and Clinical Pathology, Swedish University of Agricultural Sciences, Uppsala, Sweden. Left lateral and ventrodorsal (VD) views were obtained from unsedated dogs. The radiographs were selected from examinations collected during a 10-year period, which involved several studies on progression and treatment of myxomatous mitral valve disease in cavalier King Charles spaniels.14–19 All dogs were mature, ranging in age from 1 to 12 years. The clinical evaluation of all dogs included physical examination, auscultation of the heart, electrocardiography (ECG), thoracic radiology, and echocardiography in that order. In all dogs with cardiomegaly, echocardiography confirmed the cause to be mitral regurgitation attributable to myxomatous mitral valve disease. To be included in the study, the radiographs had to be of good technical and diagnostic quality. The basis for the selection of radiographs was to create a test situation such that the results would not be biased by great difficulties in
finding the anatomic landmarks, by obliquity or uncommon variations of thoracic conformation. The radiographic appearance had to be such that the images could be assigned into one of the five following groups: (Normal) normal cardio-pulmonary structures, (I) slight left atrial enlargement and slight cardiomegaly, (II) moderate left atrial enlargement and moderate cardiomegaly, (II þ ) moderate left atrial enlargement, moderate cardiomegaly and congestive heart failure, (III þ ) severe left atrial enlargement, severe cardiomegaly, and congestive heart failure (Table 1). Ten sets of radiographs were selected for each category. Two of the authors (K.H. and P.L.) made the selection. Each set of radiographs, consisting of one left lateral and one VD view, were assigned a randomized number between one and 50 and sorted according to number.
Observers The observers represented four levels of experience with four individuals per level. The experience levels were: (1) European Diplomates in Veterinary Diagnostic Imaging (DipECVDI), (2) Clinicians with 415 years experience of small animal clinical practice, (3) Clinicians enrolled in a 3 year national (Swedish) training program towards specialization in canine and feline diseases, all in their 2nd or 3rd year of training (trainees), (4) 5th year veterinary students whose small animal teaching was completed after the 4th year. The 16 observers evaluated the radiographs following identical oral and written instructions presented to each observer immediately before the evaluation. The observers were informed that radiographs were obtained from a
Table 1. Radiologic Inclusion Criteria for Slight, Moderate and Severe Cardiomegaly and Congestive Heart Failure I Left atrium
Overall cardiac appearance
Vascular structures and pulmonary parenchyma
II
Straight caudal border Straight caudal border or slight concavity at the level of atrio-ventricular junction Dorsal deviation and slight Minimal dorsal deviation of compression of left main stem left main stem bronchus on bronchus lateral view Normal VD view With or without bulging of left auricle on VD view Increased width of ventricular Trachea dorsally displaced but not area with a rounded apex more than parallel to the spine on on lateral and VD view lateral view Increased width of ventricular area with a generally rounded appearance on both lateral and VD view Normal Normal
II þ Straight caudal border
III þ Obvious dorsal deviation and compression of left main stem bronchus on lateral view and Obvious bulging of left auricle on VD view
Dorsal deviation and slight compression of left main stem bronchus With or without bulging of left auricle on VD view Trachea dorsally displaced but Trachea dorsally displaced, not more than parallel to towards the spine on lateral the spine on lateral view view Increased width of ventricular Heart silhouette occupies the majority of the thoracic area with a generally cavity on both lateral and rounded appearance on VD view both lateral and VD view Venous congestion and/or Venous congestion and/or non-structured increased diffuse non-structured interstitial opacity mainly in mainly perihilar increased the caudal lung lobes interstitial opacity
I ¼ slight left atrial and slight cardiomegaly; II ¼ moderate left atrial and moderate cardiomegaly; II þ ¼ moderate left atrial, moderate cardiomegaly and congestive heart failure; III þ ¼ severe left atrial, severe cardiomegaly and congestive heart failure. VD, ventro-dorsal view.
124
Hansson et al.
2005
population of dogs with varying degrees of mitral valve disease, ranging from no disease to severe disease, but were not aware of the disease status of any individual animal or the proportion of normal to diseased dogs. Each person evaluated the radiographs sequentially and at one sitting. The evaluation took 3–5 h.
Instructions for Radiographic Measurements The interpreters were instructed to measure the heart using the lateral view (Fig. 1A). The measurements were performed according to the VHS method by Buchanan, with a clarification as to which structure that should be used as the reference point at the heart base (A) and a modification of the short axis measurement (B). (A) The cardiac long axis was to be measured, in mm, with a ruler from the base of the heart to the apex. The observers were instructed to use the ventral border of the largest of the main stem bronchi seen in cross section as the reference point at the heart base. This was assumed to represent the end-on projection of the left cranial lobe bronchus before the bifurcation to the cranial and caudal lobe segments. If there were several stem bronchi of equal size, the most cranial was to be used. The same measurement points were to be used regardless of whether the left atrium was enlarged or not. (B) The cardiac short axis was to be measured, in mm, perpendicular to the length measurement. The caudal reference point for width was halfway between dorsal and ventral border of the caudal vena cava in all dogs regardless of whether the left atrium was enlarged or not. The cardiac long and short axes were transformed from mm into whole and 0.1 of VHS units by the observers. The observers were instructed that one VHS unit should include the length of a vertebral body and its caudal disc (Fig. 1B). Observers transferred the measurements of each axis to the vertebral silhouettes and measured each axis (in vertebrae to the nearest 0.1 v) from the cranio-ventral margin of T4 caudally. The two VHS measurements (for long and short axis) were then summed after the study by the first author (K.H.). In addition to the axes for the VHS, all observers measured the distance between the main stem bronchus and the spine (Br–Sp) on the lateral view (Fig. 1A and B). This measurement was to be used as an indicator of whether variations in long axis measurements arose from variation in identifying the reference point at the heart base or the cardiac apex. For the Br–Sp measurement the same reference point at the base of the heart as described above was to be used. The second reference point was an extrapolated line between the cranio-ventral and caudo-ventral surface of the vertebrae dorsal to the heart base (usually T5) (Fig. 1B). Additionally, the long axis of the fourth thoracic vertebrae (T4) including its caudal disc was measured in mm by
Fig. 1. (A) Linear dimensions and reference points for cardiac size to be used for translation into vertebral heart size (VHS). Cardiac long axis (L) is measured from the ventral border of the left cranial lobe bronchus (seen endon) to the apex. Cardiac short axis (S) is measured perpendicular to the measurement of the long axis with the caudal reference point halfway between dorsal and ventral border of the caudal vena cava. Bronchus-spine (Br–Sp) distance is measured from the ventral border of the left cranial lobe bronchus to the ventral border of the vertebra dorsal to the heart base. (B) One vertebral unit (v) and the second reference point () for bronchus to spine distance (Br–Sp) are illustrated (bold arrows). One vertebral unit is the vertebral body and its caudal disc. The dorsal reference point for Br–Sp is where the vertical measurement from the heart base intersects an imaginary line between the cranio-ventral and caudo-ventral surface. T4 ¼ fourth thoracic vertebrae.
all observers and used as an indicator of vertebral length in the thoracic spine (Fig. 1B). No measurements were done on the VD view. The view was only used in the classification of the radiographs into one of the five groups.
Statistical Analyses Statistical analyses were performed with a computerized statistical software package. For all VHS measurements, the mean, the standard deviation (SD) and the CV were calculated. The CV is deJMP 4.02, SAS Institute Inc. Cary, NC.
Vol. 46, No. 2
125
InterobserverVariability, Heart Size by VHS
Table 2. Mean, Standard Deviation, Range of Mean VHS and Coefficient of Variation (CV), Initially Calculated Per Set of Radiographs. Experience of observers
Normal 10 rads
All observers (N ¼ 16)
DipECVDI (N ¼ 4)
Experienced small animal clinicians (N ¼ 4)
Trainees (N ¼ 4)
Trainees without observer 11 (N ¼ 3)
Students (N ¼ 4)
10.8 0.49 (10.0–11.4) CV 2.5% (1.7–3.2) 10.7 0.51 (9.9–11.3) CV 2.0% (1.2–3.3) 10.8 0.52 (10.0–11.5) CV 2.3% (0.7–3.9) 10.7 0.50 (9.8–11.4) CV 3.2% (1.8–6.6) 10.9 0.52 (10.0–11.5) CV 2.0% (0.5–3.3) 10.8 0.47 (10.2–11.6) CV 2.1% (0.6–3.3)
I 10 rads
II 10 rads
II þ 10 rads
III þ 10 rads
11.3 0.35 (10.9–11.7) CV 2.6% (1.8–3.7) 11.2 0.38 (10.8–11.8) CV 2.2% (0.9–3.6) 11.3 0.36 (10.8–11.8) CV 1.9% (0.7–3.0) 11.2 0.30 (10.8–11.6) CV 3.1% (1.5–7.8) 11.3 0.34 (10.8–11.8) CV 1.5% (0.0–3.6) 11.3 0.43 (10.8–11.8) CV 2.4% (0.9–4.3)
12.3 0.62 (11.8–13.9) CV 2.0% (1.2–2.6) 12.2 0.67 (11.6–14.0) CV 1.5% (0.7–2.1) 12.4 0.62 (11.9–14.0) CV 1.8% (0.9–3.9) 12.3 0.64 (11.7–14.0) CV 2.4% (1.1–5.1) 12.4 0.66 (12.0–14.2) CV 1.4% (0.0–3.1) 12.3 0.60 (11.8–13.8) CV 1.7% (0.4–3.3)
13.0 0.36 (12.3–13.6) CV 2.0% (1.3–2.9) 13.0 0.41 (12.1–13.5) CV 1.6% (0.4–2.6) 13.1 0.34 (12.4–13.5) CV 1.7% (0.6–3.1) 12.9 0.36 (12.2–13.5) CV 2.0% (0.7–5.3) 13.0 0.35 (12.3–13.6) CV 1.8% (0.4–5.1) 13.1 0.38 (12.5–13.8) CV 1.9% (0.7–4.3)
13.8 0.65 (13.0–15.2) CV 2.4% (1.6–3.1) 13.8 0.63 (12.9–15.1) CV 1.9% (1.0–2.7) 13.9 0.64 (13.1–15.4) CV 1.9% (0.9–4.1) 13.7 0.67 (12.9–15.0) CV 2.9% (1.1–4.3) 13.9 0.69 (12.9–15.3) CV 1.5% (0.4–2.6) 13.9 0.69 (13.0–15.4) CV 2.4% (1.3–3.5)
I ¼ slight left atrial and slight cardiomegaly; II ¼ moderate left atrial and moderate cardiomegaly; II þ ¼ moderate left atrial, moderate cardiomegaly and congestive heart failure; III þ ¼ severe left atrial, severe cardiomegaly and congestive heart failure. Rads, radiographs, all radiographs are left lateral views; N, number of observers; DipECVDI, Diplomates of European College of Veterinary Diagnostic Imaging; VHS, vertebral heart size.All values are express as whole and decimal VHS units (v). For the group of Trainees, values are shown with and without (in italics) an observer who systematically measured a lower VHS compared to all other observers.
fined as the SD divided by the mean value. Initially, a radiograph-specific VHS and CV were calculated using measurements from all 16 observers (top row, Table 2). Values for each group of observers and each group of radiographs (normal to severe cardiomegaly) were calculated in the same way (Table 2). A randomized complete block analysis of variance (ANOVA) was performed with set of radiograph as blocking factor over observer as independent variable to evaluate the contribution to the overall variation. Values are reported as mean and 95% confidence interval for the individual observer (Fig. 2). Mean values for cardiac long axis (Fig. 3) and Br–Sp, in mm, were determined for each observer separately. Furthermore, the mean difference and range of difference were determined for each set of films in each group of radiographs and each group of observers (Fig. 4).
Results The overall mean VHS for the five groups of radiographs ranged from 10.8 v, in normal hearts, to 13.8 v in severe cardiomegaly (Table 2). The CV for all 16 observers was between 2.0% and 2.6% with lowest value for moderate enlarged hearts and highest value for slightly enlarged hearts. The largest difference in mean VHS value among
the groups of radiographs was between mild (group I) and moderate (group II) cardiomegaly with no overlap in VHS range between these two groups.
Variation Among Observers and Within the Levels of Experience Based on the randomized complete block ANOVA, observer had significant effect on the VHS value. The CV range was between 1.5% and 3.2% for all experience levels and all groups of radiographs. In all groups of radiographs, the highest mean and largest CV range was seen for the trainees (Table 2). However, the larger variation was because of one observer consistently measuring a lower VHS, than the other observers (Fig. 2). This interpreter measured a shorter cardiac long axis (Fig. 3) and a longer Br–Sp distance compared with all other observers. If the diverging observer was excluded, all CV values (mean and range) for the trainees were comparable with the other observer groups. The mean difference among the 16 observers when comparing the VHS values for each of the 50 sets of radiographs was 1.0 0.3 v. The smallest difference was 0.5 v and the largest 1.9 v. Excluding the single observer who systematically measured a lower VHS, the mean difference
126
2005
Hansson et al.
12.6 12.5 12.4 12.3 12.2 12.1 12 11.9 11.8 11.7 11.6 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Observer
Fig. 2. Mean (E) and 95% confidence interval for each individual observer, obtained from the randomized complete block ANOVA. One observer (number 11) measured a significantly lower vertebral heart size value. All values are corrected for the variation in heart size among the 50 sets of radiographs. Observers 1–4 ¼ European Diplomates in Veterinary Diagnostic Imaging, observers 5–8 ¼ clinicians with 415 years experience of small animal clinical practice, observers 9–12 ¼ trainees in small animal clinical practice, observers 13–16 ¼ veterinary students.
was 0.9 0.2 v and the largest difference among the observers was 1.3 v. When the radiographs were separated into different groups (normal to severe cardiomegaly), the mean difference was approximately 1 VHS unit for all groups (Fig. 4). Further separation into experience levels resulted in a mean difference among the four observers of approximately 0.5 v except for the trainees who had a higher interobserver difference for all groups of radiographs because of the diverging observer (Fig. 4). There was a tendency towards
better agreement among the DipECVDI compared with the other groups of observers.
Variation Among Experience Levels The agreement in mean VHS among the four experience levels in each group of radiographs was almost identical (Table 2). For radiographs with normal cardiac size and radiographs with slight cardiomegaly the mean VHS value
103 102 101 Cardiac long axis (mm)
100 99 98 97 96 95 94 93 92 0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Observer Fig. 3. Scatter plot of mean cardiac long axis for individual observers. One observer (number 11) measured a shorter cardiac long axis compared with others. Observers 1–4 ¼ European Diplomates in Veterinary Diagnostic Imaging, observers 5–8 ¼ clinicians with 415 years experience of small animal clinical practice, observers 9–12 ¼ trainees in small animal clinical practice, observers 13–16 ¼ veterinary students.
Vol. 46, No. 2
127
InterobserverVariability, Heart Size by VHS
Disagreement in VHS units
2.5
2
1.5
1
0.5
0 0
1
2
3
4
All
1
Normal
2
3
4
All
I
1
2
3
II
4
All
1
2
3
4
All
1
II+
2
3
4
All
III+
Fig. 4. Bar chart illustrating the mean (E), minimum and maximum difference in vertebral heart size value among the four observers in each observer group and difference among all 16 observers for each of the five groups of radiographs (10 sets of films/group of radiographs). The mean difference among all 16 observers was approximately 1.0 v. The mean difference among the observers in each observer group is approximately 0.5 v except for the trainees who had a higher interobserver difference, 0.7–0.9 v in all but one group of radiographs. There was a tendency towards better agreement among the DipECVDI compared with the other groups of observers. Observer groups: (1) European Diplomates in Veterinary Diagnostic Imaging, (2) Clinicians with 415 years experience of small animal clinical practice, (3) Trainees in small animal clinical practice, (4) Veterinary students, (All) All values for the 16 observers. Radiographs: I ¼ slight left atrial and slight cardiomegaly, II ¼ moderate left atrial and moderate cardiomegaly, II þ ¼ moderate left atrial, moderate cardiomegaly and congestive heart failure III þ ¼ severe left atrial, severe cardiomegaly and congestive heart failure.
was the same or differed by 0.1 v. For the other three groups of radiographs the difference was 0.2 v (Table 2). The lowest difference in CV values among the experience levels was 0.4% and the highest 1.2% (Table 2). The lowest CV values were found in the moderately enlarged hearts for all groups of observers (Table 2).
Variation in Measurements of Long and Short Cardiac Axes The mean VHS for normal cavalier King Charles spaniels was 10.8 0.49 v (10.0–11.4) with a CV of 2.7% (Table 2). The long axis was 5.6 0.37 v and short axis was 5.1 0.28 v. The long axis CV was 4.0% and the short axis CV was 2.2%. For all groups of radiographs (normal to severe cardiomegaly), the CV of the long axis dimension was higher than that of the short axis. The mean long axis ranged from 82.5 7.6 mm (normal hearts) to 109.7 9.5 mm (severe cardiomegaly) and the short axis from 74.9 5.7 mm (normal hearts) to 102.6 9.2 mm (severe cardiomegaly). The mean difference among the observers for measurements obtained from the same radiograph was 13.3 5.6 mm in the long axis dimension and 5.8 2.8 mm in the short axis dimension. If the observer measuring a shorter long axis was excluded the mean difference for long axis was 10.8 3.6 mm.
Variation in Measurements of the Fourth Thoracic Vertebrae The mean length of T4 was 14.6 1.0 mm (9–17). The wide range, with 9 mm as lowest value was because of two
outliers. Excluding these two values did not change the mean but decreased the range to 11–17 mm. One dog had a markedly shorter T4 with a mean length of 12.0 0.4 mm (11–13). This dog had a moderately enlarged heart by subjective classification (Table 1) but VHS (13.9 0.4 v) in the range of the severely enlarged hearts (Table 2). If this dog was excluded from the group of moderately enlarged hearts the mean VHS decreased from 12.3 to 12.1 v and upper value of the VHS range decreased from 13.9 to 12.5 v.
Discussion In this study 16 observers representing four levels of experience in interpretation of thoracic radiographs measured the heart according to the VHS method in normal dogs and dogs with varying degrees of cardiomegaly. The use of one breed eliminated variations that could be caused by breed. The CV from 1.5% to 3.2% was in the same range as previous studies in dogs (2.7–2.8%)12 and in cats (4.0%)20 for all groups of radiographs and all levels of experience. The major distinction between our study and previous studies using VHS is that the interobserver comparisons are between both individual observers and among groups of observers with different experience in interpreting thoracic radiographs. In this study, there was little variation in mean VHS values among the four groups of observers despite very different experience. For the two groups of radiographs with normal cardiac size and slight cardiomegaly the mean VHS value was the same among the observer groups or differed with only 0.1 v. In the other three groups
128
Hansson et al.
of radiographs the largest mean VHS difference among the observer groups was 0.2 v. To convert the magnitude of 0.1 v in mm one can use the mean value of T4 (14.6 mm) in this study as a guide and 0.1 v would then correspond to approximately 1.5 mm. Additionally, we analyzed the differences in VHS value measured on the same radiograph among all 16 observers and among the four observers in each experience level. When VHS values for the same radiograph were compared, the mean difference among all 16 observers was approximately 1.0 VHS unit regardless of the size of the heart. If the observer that systematically measured a shorter long axis was excluded the mean difference among the observers decreased with 0.1 VHS units. This is somewhat higher than reported by Lamb who found that the maximum likely difference (mean difference 1.96 SD) of 41.0 v among the three observers performing the measurements.13 The higher difference in our material is probably a reflection of including 16 different observers compared with three observers in the study by Lamb.13 These differences are surprisingly high and have a high impact on the use of normal VHS values and on sequential measurements on the same dog. It indicates that normal values for VHS should be used only as guidelines and not as cut off values to separate between normal and enlarged hearts. It is crucial that there is a common agreement on how measurements should be performed amongst clinicians monitoring progressive cardiomegaly in the same dog. The optimal situation is that the same clinician measures all follow up radiographs in a specific case. The mean difference in VHS among the four observers in each group was approximately 0.5–0.6 v in all groups except for the trainees where the interobserver difference was higher and sometimes as high as 0.9 v. The higher range was because of one observer who diverged from all others by measuring a shorter cardiac long axis, with resulting lower VHS values both for the long axis and for the total VHS. The same observer also measured a longer Br– Sp. The most likely explanation is that the observer used a more ventral structure than the recommended bronchus as a reference point for the heart base.
Cardiac Reference Points We decided to use a reference point on a bronchus seen end-on as it is easy to define and can be used in all dogs, regardless of cardiac size. In the original work by Buchanan the long axis was measured from the ventral border of the left main stem bronchus and according to the figure it seems that the ventral border of an end on bronchus is being used. We suspect it is the same reference point as the one we are using but it might also be slightly different from that in Buchanan’s original study.10 Our study also differs from another and later study by Buchanan that
2005
included dogs with left atrial enlargement where the long axis was measured caudal to the tracheal bifurcation to be sure of including left atrial enlargement21. If the observers in our study were to use two different reference points this would have introduced a subjective evaluation of whether the dog had cardiomegaly or not. However, it is our impression that the long axis measurement would be almost the same in many dogs regardless of which method that were being used as the heart base is semicircular and the long axis can be regarded as a radius with the apex as the center point. We found differences in consistency between measurements of both long and short cardiac axes. The higher CV value for the long axis dimension reflects more variability in the definition of the reference points. Even if the observer that systematically measured a shorter long axis was excluded the difference among observers for long axis measurements was still approximately two times that for the short axis. Possible explanations are that the reference point at the heart base can be difficult to define because of greater complexity and variability of the topographic anatomy compared with other reference points. Another difficulty might have been the presence of pulmonary edema and increased opacity in the hilar region and for this reason we elected to have two groups of radiographs with moderate cardiomegaly, one with and one without pulmonary edema. Pulmonary edema did not increased variation in long axis measurements. The measurement of Br–Sp distance was used as a crude indication of how the reference point at the heart base was selected. One observer measured a longer distance compared with others indicating that this person systematically selected a reference point at the heart base that was more ventral. The selection of the cardiac apex reference point is another potential cause of variation. Difficulties in defining the apex can be because of small amounts of pleural effusion or superimposition of ribs, skin folds or cranial parts of the liver in dogs with severe cardiomegaly. Some differences are to be expected but the magnitude was larger than expected, considering that all observers were given both oral and written instructions on how to perform the measurements, and that the written instructions were available to them throughout their evaluation of the radiographs.
Cardiac Size The mean VHS value in the normal dogs was 10.8 0.5 v, which is slightly above the suggested upper limit for normal heart size in most breeds.10 It is also approximately one vertebral unit higher than the suggested mean value 9.7 0.5 v.10 However, in the study by Lamb the normal value for cavalier King Charles spaniels was 10.6 0.5 v,11
Vol. 46, No. 2
InterobserverVariability, Heart Size by VHS
which is consistent with our value. The number, and thus contribution, of each breed comprising the original 100 dogs that were used to derive the average was not stated, and this could well influenced the stated values.10 The use of breed-specific VHS values is needed for the VHS method to have a high specificity for normal heart size. It was not the purpose of this study to establish reference ranges. Assigning the 50 sets of radiographs into five different groups was only a way to categorize the material for this study. We included radiographs with various heart sizes to cover the range encountered in practice. When comparing the VHS range for all observations and the different groups of radiographs there was an overlap between normal and slight cardiomegaly but no overlap between slight and moderate cardiomegaly. This is to be expected as dogs classified as slight cardiomegaly only had an increased cardiac width and no dorsal displacement of the trachea (Table 1). Dogs classified to have moderate and severe cardiomegaly had an increase in both long and short axis dimensions, a distinct difference.
Vertebral Measurements In our study one vertebral unit consists of both the length of the vertebrae and its caudal disc, following Buchanan’s original study for each whole number of vertebrae, but not for the decimal number. We find it more consistent to include the disc both for the decimal and whole number of vertebrae. A cardiac measurement would otherwise get the same VHS value if it reaches either the caudal endplate of one vertebra or the cranial endplate of the following vertebra. This minor divergence from the original study would not alter the interobserver results in our study but means that the decimal values in our study are slightly lower than if we had excluded the disc. This discrepancy between our VHS values and values obtained according to Buchanan’s original method is minor as a normal disc is 1.5–2 mm in width. Additionally, we wanted to clearly define the vertebral measurement points and instructed the observers to use the cranioventral and caudoventral borders. This clarification was needed because of variations in shape of the cranial vertebral endplate. Some dogs have a pronounced concavity in the mid part of the endplate, both cranial and caudal. If some observers used the mid part and some the ventral part there would be a difference in measurement data based on anatomy and not experience level. In Buchanan’s original report, using dogs with a wide range of body weights, heart size measured by the VHS method correlated with a body length index represented by two times the distance between the cranial aspect of T4 and the caudal aspect of T8. In our study, we wanted to eval-
129
uate how much the length of one specific vertebrae differed among individuals of the same breed. This can cause both intra- and interbreed variations and affect the reference value for normal dogs. For practical reasons we used T4 as a reference as this vertebrae is the cranial reference point for the VHS measurement. The length of T4 and its caudal disc varied between 11 and 17 mm if two outlier values were excluded. Hypothetically, and in a worst case scenario, this means that by using the mean values for the dimensions of normal dogs in this study, long axis 82 mm and short axis 75 mm, the VHS could range between 14.3 and 9.2 v because of variations in the vertebral lengths in individual dogs. This large variation in normal VHS values is not likely to be common but illustrates that when using the VHS method the observer also needs to examine the vertebral column for narrow discs and changes in vertebral length. In our study there was one dog, which had a considerably shorter T4 than the others and narrow disc spaces from T4 to T11. The dog had a moderately enlarged heart by our subjective classification but its VHS was in the range of the severely enlarged hearts. This dog is the reason for the wide VHS range with a high upper limit seen in the group of radiographs with moderately enlarged hearts without congestive heart failure (Table 2).
Measurement Errors Other sources of variation are mistakes such as writing the wrong numbers on the protocol, misinterpretation of the number on the ruler and the transformation of the mm measurement into VHS units. The significance of these errors in our study is uncertain. To have exact knowledge on how various observers performed measurements marks would have to have been made on each radiograph. For practical reasons this was not done.
Conclusions The VHS values for heart size can be affected by several factors. Individual variations in actual heart size and vertebral length between dogs need to be considered as well as narrowed disc spaces. The observer performing the measurements is also a cause for variation. It is shown that the VHS method for heart size is independent of observer experience but dependent of individual observers’ selection of reference points and the transformation of long and short axis dimensions into VHS units. The mean difference among observers in this study was approximately one vertebral unit. ACKNOWLEDGEMENT
The authors wishes to express their appreciation for all the work done by all observers in study.
130
Hansson et al.
2005
REFERENCES 1. Silverman S, Suter PF. Influence of inspiration and expiration on canine thoracic radiographs. J Am Vet Med Assoc 1975;166:502–510. 2. Toal RL, Losonsky JM, Coulter DB, DeNovellis R. Influence of cardiac cycle on the radiographic appearance of the feline heart. Vet Radiol 1985;26:63–69. 3. Toombs JP, Ogburn PN. Evaluating canine cardiovascular silhouettes: radiographic methods and normal radiographic anatomy. Compend Contin Educ Pract Vet 1985;7:579–587. 4. Suter PF, Lord PF. Thoracic radiography: a text atlas of thoracic diseases of the dog and cat. Wettswil: Selbstverlag PF Suter, 1984;24–34. 5. Lord PF, Suter PF. Textbook of canine and feline cardiology. Philadelphia: WB Saunders, 1999;107–129. 6. Hamlin RL. Analysis of the cardiac silhouette in dorsoventral radiographs from dogs with heart disease. J Am Vet Med Assoc 1968;153: 1444–1460. 7. Uhlig K, Werner J. Eine roentgenographische methode zur messung der herzvergroesserung beim hund. Berl Munch Tierurztl Wochenschr 1969;6:110–112. 8. Buchanan JW. Radiology of the heart. Proceedings of the 35th Annual Meeting of American Animal Hospital Association, 1968; 35–36. 9. Kealy JK. Diagnostic radiology of the dog and cat. Philadelphia: WB Saunders, 1979;215. 10. Buchanan JW, Bu¨cheler J. Vertebral scale system to measure canine heart size in radiographs. J Am Vet Med Assoc 1995;206:194–199. 11. Lamb CR, Wikeley H, Boswood A, Pfeiffer DU. Use of breedspecific ranges for the vertebral heart scale as an aid to the radiographic diagnosis of cardiac disease in dogs. Vet Rec 2001;148:707–711. 12. Nakayama H, Nakayama T, Hamlin RL. Correlation of cardiac enlargement as assessed by vertebral heart size and echocardiographic and electrocardiographic findings in dogs with evolving cardiomegaly due to rapid ventricular pacing. J Vet Intern Med 2001;15:217–221.
13. Lamb CR, Tyler M, Boswood A, Skelly BJ, Cain M. Assessment of the value of the vertebral heart scale in the radiographic diagnosis of cardiac disease in dogs. Vet Rec 2000;146:687–690. 14. Ha¨ggstro¨m J, Hansson K, Karlberg BE, Kvart C, Olsson K. Plasma concentration of atrial natriuretic peptide in relation to severity of mitral regurgitation in Cavalier King Charles Spaniels. Am J Vet Res 1994;55:698– 703. 15. Ha¨ggstro¨m J, Hansson K, Karlberg BE, Kvart C, Madej A, Olsson K. Effects of long-term treatment with enalapril or hydralazine on the reninangiotensin-aldosterone system and fluid balance in dogs with naturally acquired mitral valve regurgitation. Am J Vet Res 1996;57:1645–1652. 16. Ha¨ggstro¨m J, Hansson K, Kvart C, Karlberg BE, Vuolteenaho O, Olsson K. Effects of naturally acquired decompensated mitral valve regurgitation on renin-angiotensin–aldosterone system and natriuretic peptide concentration in dogs. Am J Vet Res 1997;58:77–82. 17. Ha¨ggstro¨m J, Hansson K, Kvart C, Vuolteenaho O, Olsson K. Relationship between different natriuretic peptides and severity of naturally acquired mitral regurgitation in dogs with myxomatous valve disease. J Vet Cardiol 2000;1:7–16. 18. Kvart C, Ha¨ggstro¨m J, Duelund Pedersen H, Hansson K, et al. Efficacy of enalapril for prevention of congestive heart failure in dogs with myxomatous valve disease and asymptomatic mitral regurgitation. J Vet Intern Med 2002;16:80–88. 19. Hansson K, Ha¨ggstro¨m J, Kvart C, Lord P. Left atrial to aortic root indices using two-dimensional and M-mode echocardiography in Cavalier King Charles spaniels with and without left atrial enlargement. J Vet Radiol Ultrasound 2002;43:568–575. 20. Litster AL, Buchanan JW. Vertebral scale system to measure heart size in radiographs of cats. J Am Vet Med Assoc 2000;216:210–214. 21. Buchanan JW. Vertebral scale for canine and feline heart size. Vet Clin North Am Small Anim Pract 2000;30:379–393.
