A total of 28 healthy mares owned by–masked for review–were included in the study that took place between January and June 2020. The mares were considered healthy based on history, physical examination, ECG and echocardiography. The age ranged between 4 and 20 years old (median 8.5 years), the BW between 430 and 630 kg (median 522 kg) and the BCS score between 4 and 5/9 (median 4.5). They were 21/28 Standardbred (75%), 6/28 Thoroughbred (21%), and one Anglo-Arab (4%). The research protocol was approved by the Institutional animal care and use committee of the University of Pisa (45965/2016).

For the study protocol each mare was restrained in a stock and, after 5 minutes of acclimatization period, two recordings of 30 seconds were realized simultaneously with the rECG and the spECG.

The rECG (Televet 100, Engel Engineering Service GmbH, Germany) was acquired using surface electrodes attached to the skin with alligator clips using a standard base-apex system [13]. The yellow positive electrode (left arm) was positioned on the left precordial area, the red negative electrode (right arm) was positioned on the right jugular groove in the middle third of the neck, and the neutral and ground electrodes (black and green) was placed at the tip of the left shoulder and at the level of the sternum, respectively. Alcohol was applied on the clips to ensure electrical contact. The ECG tracings were printed with a paper speed of 25 mm/s with an amplitude of 10 mm/mV. A sampling rate of 500 Hz was used for standard ECG acquisition, with a 100 Hz low-pass filter and a 0.05 Hz high-pass filter.

The spECG device was applied using a single-lead bipolar ECG (1-lead ECG eKuore, Chip Ideas Electronics SL, Spain) and its smartphone application (eKuore ECG, Chip Ideas Electronics SL, Spain) (Fig 1).

The tracings were recorded with an iPhone 5S (Apple, USA), automatically digitalized by the device, archived as PDF and printed with a paper speed of 25 mm/s and an amplitude of 10 mm/mV. The two electrodes were applied with two different methods previously described: a standard (SBA) and a modified (MBA) base-apex derivation [12,14]. A first 30 seconds-recording was obtained with the rECG and the SBA configuration and, immediately after repositioning the electrodes of the spECG, another 30 seconds-recording was obtained again with both rECG and the MBA configuration of the spECG. For the SBA position the positive electrode was placed on the left precordial area, and the negative electrode was placed on the right jugular groove in the middle third of the neck (Fig 2).

For the MBA position the positive electrode was maintained in the same position, while the negative electrode was applied on the left side of the withers (Fig 3).

Alcohol was applied on the skin before the placement of the electrodes to ensure electrical contact. The same operator (CF) was responsible of the placement and recordings of both the rECG and spECG in all horses.

All smartphone ECG tracings were evaluated by an experienced operator (TV) blinded to the study to assess whether the tracings were acceptable or not for interpretations. Tracings were considered acceptable for interpretation if baseline artifacts were absent for at least the 80% of the recording. Baseline artifacts were defined as ECG segments in which P-waves and/or QRS complexes could not be identified. The same operator performed also the ECG measurements on all tracings, using lead I of the rECG and the only available lead of the spECG. The following variables were measured for both rECG and spECG in each recording session: mean heart rate (HR; bpm), P-wave duration (milliseconds [ms]), PQ interval (ms), QRS complex duration (ms), and QT interval (ms). Heart rate (HR) was calculated counting the number of QRS complexes on 15 cm and multiplying the result by 10 for the spECG (paper speed of 25 mm/s). Mean HR was calculated as the mean value of three independent HR calculations from three different areas on the ECG tracings. The duration of each wave or interval was calculated as the average of the measurements of three randomly selected heartbeats.

The polarity of P-waves and QRS complexes, as well as the presence/absence and duration (ms) of artifacts were also evaluated.

Analysis was performed only with paired ECG tracings that were considered acceptable for interpretation and the rECG was considered the reference method. Cohen’s k test was used to calculate the agreement between the spECG and the rECG for the classification of the polarity of P-waves (positive, negative or biphasic) and QRS complexes (positive or negative). The k coefficient was interpreted as follows: k < 0.01: null agreement; 0.01< k < 0.20: poor agreement; 0.21 < k < 0.40: modest agreement; 0.41 < k < 0.60: moderate agreement; 0.61 < k < 0.80: good agreement; 0.81 < k < 1.00: excellent agreement. If the contingency table reported one or more values equal to zero, Cohen’s k could not be calculated and therefore in these cases the percentage of agreement was used instead.

Differences in the prevalence of baseline artifacts on spECG and rECG were evaluated using McNemar test. This test compares the observed frequencies (presence/absence of artifacts on spECG vs rECG) with those expected.

Wilcoxon matched-pairs signed ranks test for non-paired data was used to verify the differences in number, duration and prevalence of artifacts. The Bland Altman test was used to evaluate the differences in duration of the electrical defections with spECG and rECG and 95% bias and limits of agreement were calculated for HR, P-wave duration, PR interval, QRS complex duration and QT interval.

Statistical analysis was performed with a commercial software (Microsoft Excel, 2011; GraphPad Prism 6, USA) and a P value < .05 was set as significant.

Within the subjects included in the study, all 28 mares showed a normal sinus rhythm, one of them (3.6%) presented a single episode of second-degree atrioventricular block according to rECG.

Among the smartphone ECG tracings, all the 56 recordings obtained from the 28 mares with the two methods of electrodes placement were considered acceptable for interpretation, thus no one was excluded from the statistical analysis.

With the first 30 seconds of recording the median HR was 60 bpm with both the rECG and the spECG with the SBA configuration. The bias between the HR measured manually on rECG and SBA configuration of the spECG was 4 bpm (95% CI: -5 to 12 bpm). The median HR obtained with the successive 30 seconds of recording was 50 bpm with both the rECG and the MBA configuration of the spECG. The bias between the HR measured manually on rECG and spECG with MBA configuration was 1 bpm (95% CI: -5 to 3 bpm).

The median duration of the electrical deflections measured on the tracings from the rECG and the two electrodes configurations of the spECG are reported in Table 1, along with the bias calculated from their comparison.

P-wave polarity was positive in all recordings with both the rECG and spECG, independently from electrode configuration. Therefore, an excellent agreement was found when comparing the rECG and the spECG for the evaluation of the polarity of P-waves. QRS complexes had negative polarity in all recordings with both the rECG and spECG, independently from electrode configuration. Thus, the agreement between the rECG and the spECG for the assessment of the QRS complex polarity was excellent (Fig 4).