Reason for performing study: The repeatability of various echocardiographic measurements is unknown.
Objectives: To determine the intraoperator, intraobserver and interoperator variability of echocardiographic measures in healthy foals.
Methods: Echocardiographic examinations were carried out on 6 healthy foals by 3 experienced echocardiographers. Intraoperator variability was determined by having a single echocardiographer obtain and measure images from 6 foals scanned on 3 consecutive days. Interoperator and intraobserver variability were determined by having 3 echocardiographers each obtain images from an additional 6 sedated foals. Within-day interoperator variability was determined by having each echocardiographer measure their own images. Intraobserver variability was determined by having a single echocardiographer measure images obtained by all 3 echocardiographers. The coefficient of variation (CV) and standard error were calculated for each measure.
Results: The variability for most measurements was either very low (CV<5%) or low (CV = 5–15%). Measurements of right ventricular internal diameter (RVID) in systole and E-point to septal separation (EPSS) showed moderate (CV 15–25%) to high variability (CV>25%) in all 3 categories. Measurements of the left ventricular ejection time (LVET) and velocity time integral from the right parasternal long axis view of right outflow tract in the fourth intercostal space showed moderate intraoperator variability. Measurements of the LVET, RVID in diastole and left atrial appendage (LAA) showed moderate interoperator variability and measurements of the RVID in diastole and acceleration time from the short axis view of the right outflow tract in the right third intercostal space showed moderate interobserver variability.
Conclusion: The intraoperator, intraobserver and interoperator variabilities for most echocardiographic measurements in foals are low.
Potential relevance: Most standard transthoracic echocardiographic measurements in foals have a low enough variability to warrant their use in serial clinical evaluations or experimental studies. Repeated measurements of RVID, EPSS, LVET and LAA should be interpreted with caution.
While a standardised technique for foal echocardiography has been described (Stewart et al. 1984) and normal values published for a few select breeds and ages (Rovira and Munoz 2009; Collins et al. 2010), there is little information available on the repeatability of serial echocardiographic measurements in foals. Potential sources of variability in echocardiographic quantification include day-to-day or even heart beat-to-heart beat variations in cardiac chamber or vessel size (biological variability) as well as different measurement or image acquisition techniques by different echocardiographers (measurement and recording variability). These sources of variability are important to recognise and distinguish from biological changes induced by disease. In the healthy mature horse, several studies have looked at the types and magnitude of variation in echocardiographic measurements (Young and Scott 1998; Kriz and Rose 2002; Buhl et al. 2004; Helwegen et al. 2006). In general, transthoracic 2D and M-mode measurements show less intraobserver variability than Doppler derived measurements (Young and Scott 1998).
The aim of this study was to determine the variability of standard and some nonstandard 2D, M-mode and selected spectral Doppler echocardiographic measurements in healthy Thoroughbred foals under 3 conditions: 1) when a single echocardiographer obtains and measures standardised images on separate days (intraoperator, day-to-day variability); 2) when images are obtained and measured by 3 different echocardiographers on the same day (interoperator variability); and 3) when images are obtained by 3 different echocardiographers on the same day but measured by a single individual (intraobserver variability).
Materials and methods
Twelve Thoroughbred foals were studied prospectively. All foals were considered healthy on the basis of physical examination findings, cardiac auscultation, electrocardiographic and echocardiographic examination. The 6 foals used to determine interoperator and intraobserver variability were 4 fillies and 2 colts aged mean ± s.e. 63 ± 23 days (range 36–93 days) and weighing 138 ± 19 kg (range 100–148 kg). The 6 foals used to determine intraoperator variability were 3 fillies and 3 colts aged 71 ± 11 days (range 54–91 days) and weighing 146 ± 4 kg (range 141–154 kg). The study was approved by the Local Ethics Committee for the Equine Fertility Unit, Newmarket, UK and performed under licence in accordance with the Animal (Scientific Procedures 1986) Act.
Echocardiographic examinations were performed with the foal standing and restrained by experienced handlers. The foals were neither specifically trained to tolerate the examinations nor accustomed to the observers or handlers. For the study of intraobserver and interoperator variability, all 6 foals were sedated with xylazine (0.26 ± 0.07 mg/kg bwt i.v.; range 0.15–0.36 mg/kg bwt) 27 ± 14 min prior to the start of the first echocardiographic evaluation (range 2–58 min). In relation to a concurrent study, catheters had been placed in the pulmonary arteries of these foals before these examinations were performed. All 6 foals then underwent complete echocardiographic examination by 3 individuals experienced in foal echocardiography. For the study of intraoperator variability, the foals were not sedated and had not been previously catheterised. All transthoracic 2D, M-mode and spectral Doppler echocardiographic examinations were performed according to standard technique (Long et al. 1992) using a digital echocardiograph (Logic E)a and a phased array sector transducera of 3.5 MHz. A single lead electrocardiogram was recorded simultaneously. Images were obtained from both right and left parasternal imaging planes.
Nonstandard measurements included measurement of the pulmonary artery in diastole, 4 cm distal to the valve from a left parasternal long axis view of the right ventricular outflow tract imaged through the left triceps muscle by placing the transducer approximately in the middle of the triceps adjusting it in dorsoventral and craniocaudal planes to locate the acoustic window ventral to the lung within the third intercostal space (PALTRI; Fig 1, Supplementary Item S1), measurement of the pulmonary artery in diastole from the right parasternal long axis view of the left ventricular outflow tract (PARLOT4L; Fig 2, Supplementary Item S2), measurement of the pulmonary artery in diastole from the right parasternal short axis view of the right outflow tract (PARROT4S; Fig 3, Supplementary Item S3) and measurement of the pulmonary artery in diastole from the right parasternal long axis view of the right outflow tract obtained in the right third intercostal space (PAR3L; Fig 4, Supplementary Item S4). Spectral Doppler derived pulmonary velocity time integrals were obtained from the right parasternal outflow tract view (PAR4L; Fig 5) and from the right parasternal short axis view (PARROT4S; Fig 6).
To determine intraoperator variability, 6 foals underwent 3 complete echocardiographic examinations over the course of 6 days by a single experienced echocardiographer (J.S., M.M.D. or C.M.M.). The images were then analysed by the same individual. To determine intraobserver and interoperator variability, each of 6 foals underwent 3 echocardiographic examinations, once by each echocardiographer. All 3 echocardiographic examinations for each foal were performed on the same day and the operator order was randomised each day and the entire study was completed in 7 days. Interoperator measurement variability was determined by having each echocardiographer perform and measure her own scans. Intraobserver variability was determined by having a single echocardiographer (J.S., M.M.D. or C.M.M.) measure examinations obtained by each echocardiographer.
All echocardiographic measurements were performed off-line from digital recordings of still images and cine-loops. Measurements were performed in random order and with the observers blinded to previous measurements. Three nonconsecutive cardiac cycles were measured and averaged for each variable. The R-R intervals were determined for the preceding beat when measuring ejection time, fractional shortening and Doppler indices. No attempt was made to ensure that the same frames were measured by each observer.
Statistical analyses were performed using commercial computer software (SAS software, version 9.1)b. The D'Agostino-Pearson Omnibus test was used to evaluate the normality of all variables. Intraoperator, interoperator and intraobserver variability were assessed using individual, one-way repeated measures analysis of variance for each source of variability. As established in previous studies, clinical significance was assessed using the calculated coefficient of variation produced in the analysis of variance tables. The standard error for the coefficient of variation was calculated using the equation previously described (Simpson et al. 2007). The resulting standard errors can be used to produce 95% confidence intervals for the coefficient of variation (CV). In keeping with a previous study, the degree of variability was defined as follows: CV <5%, very low variability; CV 5–15%, low variability; CV 16–25% moderate variability; >25% high variability (Grenacher and Schwarzwald 2010).
The variability data for all echocardiographic measures are summarised in Tables 1 and 2. Intraoperator, interoperator and intraobserver variability were very low or low for most 2D and M-mode measures, with the exception of RVIDs and EPSS, which had moderate to high variability. Measurements of the LVET, RVIDd and LAA had moderate variability in one or 2 categories (Table 1).
Table 1. Coefficient of variation (%CV) and s.e. in the measurement of 2D and M-mode echocardiographic parameters
Intraoperator (n = 18)
Intraobserver (n = 18 )
Interoperator (n = 18)
n=the total number of measurements performed; * indicates a nonstandard measurement of the pulmonary artery. The degree of variability was defined as follows: CV<5%, very low variability; CV 5–15%, low variability; CV 16–25% moderate variability; >25% high variability.
Table 2. Coefficient of variation (%CV) and s.e. in the measurement of spectral Doppler echocardiographic parameters
Intraoperator (n = 18)
Intraobserver (n = 18)
Interoperator (n = 18)
* indicates a nonstandard measurement of the pulmonary artery; n = the total number of measurements performed. The degree of variability was defined as follows: CV<5%, very low variability; CV 5–15%, low variability; CV 16–25% moderate variability; >25% high variability.
Interoperator was very low or low for all spectral Doppler measurements, intraoperator variability was very low or low for all spectral Doppler measurements except VTI measured from the right fourth intercostal space, in the long axis view of the right outflow tract image while intraobserver variability was low or very low for most spectral Doppler measurements with the exception of AT, which was moderately variable when measured from the right fourth intercostal space, in the short axis view of the right outflow tract.
The results of this investigation indicate that intraoperator, intraobserver and within-day interoperator variabilities as defined in this study are low for most standard 2D, M-mode and select spectral Doppler measurements in healthy, standing foals. Variability was generally higher for most echocardiographic measurements once they were obtained or analysed by more than one individual and where they were performed on different days. However, even with multiple sonographers, the CV for most measurements remained under 15%.
Comparison of this study with others in the literature is difficult because of lack of standardised statistical methods or reporting terminology and variable study designs. In this study, the results were categorised based on a previously reported, albeit arbitrarily defined, scheme (Grenacher and Schwarzwald 2010). Using this scheme, the low to very low day-to-day intraoperator variability of standard 2D and left ventricular M-mode echocardiographic measurements in this study is similar to that in mature Standardbred horses (Kriz and Rose 2002) where coefficients of variation ranged from 1.1 to 6.92%. The low day-to-day intraoperator variability in this study is also similar to that for standard left ventricular M-mode measurements (CV = 1.3–3.6%; reported as intraobserver variability) (Grenacher and Schwarzwald 2010) and left atrial diameter (CV = 1.6%; reported as intraobserver variability) (Schwarzwald et al. 2007) in mature horses of various breeds. Day-to-day intraobserver variability of pulmonary artery velocity time integral and right ventricular ejection time in mature Thoroughbreds has been reported to be low (CV = 6% for each), while acceleration time is moderately variable (CV = 17%) (Young and Scott 1998). In this study, AT had less variability when measured from the long axis view of the right ventricular outflow tract.
With the exception of the intraoperator variability of RVIDs, the variability of right ventricular measurements was moderate to high. Possible technical reasons for poor repeatability when measuring the right ventricle include inconsistent orientation of the M-mode cursor through the right ventricle and poor demarcation of the endocardial borders. The first problem is probably due to the complex anatomic geometry of the right ventricle, making standardisation of the cursor placement difficult. The latter problem may result from poor near field resolution as the M-mode views are typically optimised for the left ventricle not the right. Cyclical change in venous return in association with respiration may be a biological source of variability as right ventricular diastolic and systolic dimensions are known to increase with inspiration (Norgard and Vik-Mo 1992). In mature horses, the day-to-day intraoperator variability of right ventricular measurements on M-mode has been previously described as moderate and more reliable than 2D measurements of the right ventricle (CV 8.4 and 15.9% RVIDd and RVIDs) (Helwegen et al. 2006). This is in contrast to the American Society of Echocardiography's recommendation for human patients where quantitative assessment of RV dimensions is best performed from the 2D apical long axis view (Rudski et al. 2010). Apical views are difficult to obtain in equine cases due to a limited acoustic window and were not included as part of this study.
Numerous nonstandard measurements of the pulmonary artery were obtained in this study. The views through the triceps muscle were developed because some foals will not tolerate placement of the transducer under the triceps muscle into the third intercostal space. Additionally, with the relatively low frequency transducer that was used in this study, this approach allowed visualisation of the structures of interest within the focal zone. With higher frequency transducers providing superior images in the near field, this approach may not necessary. The additional views of the pulmonary artery also reflect the authors' interest in using echocardiography in the diagnosis and management of pulmonary arterial hypertension. In foals, pulmonary arterial hypertension may appear secondary to meconium aspiration, acute respiratory distress syndrome, and persistent fetal circulation (Wilkins 2003). Studies are needed to determine if echocardiography can be reliably used to diagnose pulmonary arterial hypertension in foals and whether serial measurements of the pulmonary artery diameter can reliably identify changes in pulmonary artery pressure in response to therapy or disease progression. In this study, the low variability of the pulmonary artery measurements justifies further investigations into their use for following changes in pulmonary arterial pressures in foals.
The mitral E-point to septal separation is an M-mode measurement of the distance between the most posterior point of the interventricular septum at end diastole and the E-point of the anterior mitral valve leaflet in the same cardiac cycle and is used as an indicator of left ventricular function (Ahmadpour et al. 1983; Lehmann et al. 1983). While normal values have been determined, variability of this measurement in horses has not been reported previously. In this study, EPSS was highly variable regardless of operator or observer status. This is probably due to inconsistent orientation of the cursor through the mitral valve but may have also been affected by sedation. Regardless of the cause, the results of this study suggest that differences >24% on serial examinations must be achieved in order to document genuine change.
M-mode measurements with moderate variability in this study include LVET and LAA. Left ventricular ejection time is affected by heart rate and is therefore likely to be subject to beat-to-beat variability. This may have contributed to the intraoperator variability of LVET. Foals in this group were not sedated and therefore more likely to have significant changes in heart rate due to excitement as demonstrated by the variable R-R intervals in the beats preceding measurement of the LVET (CV = 22%). Additionally, one must be able to visualise the opening and closing of the aortic valve clearly in order to measure LVET accurately. This requires precise placement of the M-mode cursor. It is likely that inconsistent M-mode cursor alignment contributes to the variability in both LVET and LAA in this study. In contrast, LVET in mature horses has been shown to have low variability (Young and Scott 1998; Kriz and Rose 2002; Grenacher and Schwarzwald 2010).
The Doppler derived systolic time intervals of AT (time to peak PA flow velocity) and ET (right ventricular ejection time) and have been used to estimate pulmonary artery pressure in dogs (Schober and Baade 2006) and human patients (Chan et al. 1987). To our knowledge, these have not been investigated in foals or mature horses. Systolic time intervals, which were not assessed in the current study, may have an advantage over measuring flow velocities as estimates of pulmonary artery pressure as they are less affected by the angle of interrogation and therefore less likely to underestimate pulmonary artery pressure. Given the difficulty in Doppler alignment with flow in horses due to limited acoustic windows, systolic time intervals should be investigated to determine if they can accurately predict pulmonary artery pressure. Subjective assessment of the short-axis image of the pulmonary artery used in the current study, suggests that optimal alignment to flow is unlikely to be achieved in this imaging plane.
It should be pointed out that this study was not designed to determine the different contributions of measurement variability and recording variability to overall interoperator variability. Instead, the specific conditions of this study were chosen to mimic clinical or research conditions where: 1) serial examinations may be performed and interpreted by the same individual; 2) examinations may be performed by one individual but interpreted by another as is the case with consultation; and 3) serial examinations are performed and interpreted by more than one individual on the same day as may be the case in a large hospital with multiple clinicians providing care. It could be argued that repeated examinations performed either on the same day over a very short period of time (intraobserver and interoperator variability) or over 6 days (intraoperator variability) have limited clinical application. However, these situations might apply when monitoring a foal's cardiovascular response to a particular intervention such as volume resuscitation or the administration of vasopressors and inotropes. This within-day interoperator variability would not reflect biological variability of the foal that can be expected to occur from day-to-day. The authors theorise that this biological variation is likely to be small in healthy foals examined over a relatively short period of time but would increase more significantly the longer the time between examinations, as growth of the foal would affect echocardiographic measurements. Further studies would need to be done to support or disprove this hypothesis.
There are several limitations of this study. First, it should be pointed out that all 3 echocardiographers are experienced in foal echocardiography. Studies in cats have shown that within-day intraobserver variability of the beginner echocardiographer is high compared with more experienced echocardiographers (Chetboul et al. 2003). As this study was not designed to investigate observer/operator experience, the findings of this study would not necessarily apply to situations where echocardiographers are of varying skill levels. Second, all of the foals in this study were examined while standing. Clinically ill foals are often recumbent and unwilling or unable to stand for a complete echocardiographic examination. Studies in dogs have shown that positioning (standing vs. lateral recumbency) does not affect the repeatability of left ventricular M-mode measurements (Chetboul et al. 2005). Alternate positioning of the foal was not investigated in this study and results cannot necessarily be directly applied. Third, some of the foals in this study received a wide range of sedation over variable time periods before echocardiographic evaluation. This is a clinically relevant issue as foals are variable in their tolerance of the echocardiographic procedure and may require sedation at different dosages in order to complete an examination. The addition of sedation may affect both measurement and biological variability, although this was not specifically determined in this study. Finally, the same echocardiographic equipment and technical settings were used for all examinations in this study and the use of different equipment may increase variability due to differences in resolution and measurement accuracy.
In conclusion, most standard transthoracic echocardiographic measurements in foals have a low enough variability to warrant theiruse in serial clinical evaluations or experimental studies. In general, variability is less if a single operator acquires and measures serial scans. Repeated measurements of right ventricular internal diameters, EPSS, LVET, and LAA, even by a single operator, should be interpreted with caution.
Authors' declaration of interests
No conflicts of interest have been declared.
Source of funding
Beaufort Cottage Educational Trust.
The authors are grateful to Professor Twink Allen (Equine Fertility Unit, Newmarket, UK) for the provision of facilities and animals.