J Clin Monit Comput DOI 10.1007/s10877-015-9731-6

ORIGINAL RESEARCH

Comparison of electrical velocimetry and cardiac magnetic resonance imaging for the non-invasive determination of cardiac output Frederik Trinkmann1 • Manuel Berger1 • Christina Doesch1,3 • Theano Papavassiliu1,3 • Stefan O. Schoenberg2,3 • Martin Borggrefe1,3 Jens J. Kaden1 • Joachim Saur1

•

Received: 13 January 2015 / Accepted: 23 June 2015 Ó Springer Science+Business Media New York 2015

Abstract A novel algorithm of impedance cardiography referred to as electrical velocimetry (EV) has been introduced for non-invasive determination of cardiac output (CO). Previous validation studies yielded diverging results and no comparison with the non-invasive gold standard cardiac magnetic resonance imaging (CMR) has been performed. We therefore aimed to prospectively assess the accuracy and reproducibility of EV compared to CMR. 152 consecutive stable patients undergoing CMR were enrolled. EV measurements were taken twice before or after CMR in supine position and averaged over 20 s (AESCULONÒ, Osypka Medical, Berlin, Germany). Bland–Altman analysis showed insufficient agreement of EV and CMR with a mean bias of 1.2 ± 1.4 l/min (bias 23 ± 26 %, percentage error 51 %). Reproducibility was high with 0.0 ± 0.3 l/min (bias 0 ± 8 %, percentage error 15 %). Outlier analysis revealed gender, height, CO and stroke volume (SV) by CMR as independent predictors for larger variation. Stratification of COCMR in quintiles demonstrated a good agreement for low values (\4.4 l/ min) with bias increasing significantly with quintile as high & Joachim Saur [email protected] 1

1st Department of Medicine (Cardiology, Angiology, Pneumology, Intensive Care), University Medical Center Mannheim, Medical Faculty Mannheim, Heidelberg University, Theodor-Kutzer-Ufer 1-3, 68167 Mannheim, Germany

2

Institute of Clinical Radiology and Nuclear Medicine, University Medical Center Mannheim, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany

3

DZHK (German Center for Cardiovascular Research) Partner Site, University Medical Center Mannheim, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany

as 3.1 ± 1.1 l/min (p \ 0.001). Reproducibility was not affected (p = 0.71). Subgroup analysis in patients with arrhythmias (p = 0.19), changes in thoracic fluid content (p = 0.51) or left heart failure (p = 0.47) could not detect significant differences in accuracy. EV showed insufficient agreement with CMR and good reproducibility. Gender, height and increasing CO and SV were associated with increased bias while not affecting reproducibility. Therefore, absolute values should not be used interchangeably in clinical routine. EV yet may find its place for clinical application with further investigation on its trending ability pending. Keywords Cardiac magnetic resonance imaging  Cardiac output  Electrical velocimetry  Impedance cardiography  Non-invasive

1 Introduction Cardiac output (CO) is an important physiological parameter when assessing the function of the cardiovascular system. Although being potentially valuable in different clinical settings, broad application of CO measurement is often limited due to invasive procedures. Reliable non-invasive techniques would widen the spectrum of hemodynamic monitoring inside and outside the intensive care unit. Despite being non-invasive, a novel approach in modern medicine should also be accurate and precise. Therefore, cardiac magnetic resonance imaging (CMR) has become the new non-invasive gold standard for the assessment of cardiac function [1]. Nevertheless, the technique is expensive, time consuming and not commonly available.

123

J Clin Monit Comput

Impedance cardiography (ICG) has attracted wide interest being an easily applicable, non-invasive and inexpensive approach. The first description was made by Kubicek et al. in the 1960s [2]. The technique is based on the application of an alternating electrical current of constant amplitude to the thorax. Despite of initially promising results, limitations associated with a worse accuracy soon became apparent being due to changes in intrathoracic water content, arrhythmias and movement artifacts [3, 4]. Moreover, the estimation of an accurate left ventricular ejection time (LVET) from the ICG signal is essential [5]. As a consequence, many modifications have been made refining the original formula and leading to a vast amount of validation studies with varying results. One of the most recent adaptations is referred to as electrical velocimetry (EV) and incorporated in the AesculonÒ Cardiac Output Monitor (Osypka Medical, Berlin, Germany). In addition to the change of thoracic resistance it also relates the maximum rate of change of impedance to peak aortic blood acceleration. EV has already been extensively investigated with thermodilution being the most common invasive reference technique. While some trials indicated good agreement [6–11], others came to contrary results in critically ill patients [12] or patients undergoing right heart catheterization [13]. Moreover, EV does not adequately reflect pre- to postoperative changes in CO in cardiac surgery patients [14]. When being compared to other non-invasive techniques, Tomaske et al. [15] did not find a good correspondence between ICG and subxiphoidal Doppler flow measurements in children with congenital heart defects. Values obtained with inert gas rebreathing (IGR) should not be used interchangeably in the clinical setting [16]. With transthoracic echocardiography being the predominant technique in pediatric populations, EV showed a good agreement in healthy neonates as well as infants following switch operation for simple transposition of the great arteries [17]. Although most reference techniques used in previous studies suffer from inaccuracies themselves, EV has not been validated against the non-invasive gold standard CMR to date. Therefore, the aim our prospective study was to assess the accuracy and reproducibility of measurements obtained by EV compared to CMR in a large and heterogeneous patient collective.

2 Methods 2.1 Subjects The study protocol was approved by the Medical Ethics Committee II of the Medical Faculty Mannheim, University Heidelberg, Germany (2006-189N-MA, addendum) and registered at clinicaltrials.gov (NCT00945048). Written informed consent was obtained from all patients

123

following a full explanation of the purpose of the study as well as of risks and discomforts associated with participation. We enrolled 152 consecutive stable patients undergoing CMR with information on concomitant diseases given in Table 1. All participants were breathing spontaneously. Patients with severe aortic or mitral valve disease were excluded from the study since the volume based hemodynamic CMR measurements may be impaired. The valve status was assessed qualitatively based on the CMR images. 2.2 Study protocol All measurements were performed by EV and CMR experienced physicians and technicians blinded to the results obtained by the other method. EV measurements were performed immediately before or after the CMR examination. Two pairs of standard electrocardiographic electrodes each were placed at defined positions at the left base of the neck and the inferior aspect of the thorax at the level of the xiphoid process with an inter-electrode gap of 5 cm. Verification of the correct signal quality was accomplished by visualization of the ECG, the impedance waveform and its first derivative. Measurements were carried out after 10 min of rest in a stable supine position and averaged over a period of 20 s. To calculate reproducibility measurements were performed twice with an interval of 5 min. Both measurements were conducted either directly before or after CMR imaging.

Table 1 Concomitant diseases n (%) Cardiac Heart failure

76 (50)

Arterial hypertension

48 (32)

Coronary heart disease (CHD)

37 (24)

Myocardial hypertrophy

34 (22)

Myocardial infarction

27 (18)

Mitral valvulopathy

20 (13)

Dilated cardiomyopathy (DCM)

19 (13)

Aortic valvulopathy

18 (12)

Pericardial effusion

17 (11)

Atrial fibrillation (all types)

17 (11)

Tricuspid valvulopathy

13 (9)

Hypertrophic cardiomyopathy (HCM) Pulmonary Pleural effusion

8 (5) 10 (7)

Pulmonary hypertension

7 (5)

Chronic obstructive pulmonary disease (COPD)

5 (3)

Bronchial asthma

4 (3)

J Clin Monit Comput

2.3 Electrical velocimetry (EV) For measurement of CO we used the AESCULONÒ monitor (Osypka Medical, Berlin, Germany) which implements a newly developed algorithm referred to as EV. It additionally takes the change of erythrocytic orientation during the heart cycle into account. The maximum rate of change of thoracic electrical bioimpedance (TEB) is interpreted as the ohmic equivalent of mean blood flow velocity in the ascending aorta [vAORTA (s-1)]. The peak magnitude of the first derivative of TEB and the left ventricular ejection time are measured from the Z(t)/dt-curve and entered into the Bernstein–Osypka-equation as followed [18]: SV ¼ VEPT  vAORTA  FTc

ð1Þ

SV stroke volume (ml), VEPT volume of electrically participating tissue calculated as a function of body weight and height (ml), vAORTA ohmic equivalent of mean aortic blood velocity during left ventricular ejection (s-1), FTc flow time derived from LVET normalized for RR interval (s). The ohmic equivalent of mean aortic blood flow acceleration is transformed into an equivalent of mean aortic blood flow velocity in order to estimate vAORTA :  1 ðZ ðtÞ=dtÞMAX 2 vAORTA ¼ ð2Þ Z0 (Z(t)/dt)Max maximum rate of change of TEB during systole, Z0 base impedance. CO can then be calculated as the product of SV and heart rate (HR). Invalid measurements were most often due to misrecognition of the ECG R-wave caused by a prominent T-wave, extrasystoles or a deformed QRS complex. Other causes were overlay of the impedance signal by respiratory modulation or a misrecognition of the systolic maxima in the Z(t)/dt signal.

muscles were included in left ventricular cavity volumes [19]. For volume determination the areas subtended by the endocardial tracings were manually determined in each end-diastolic and end-systolic slice and multiplied by slice thickness to yield the end-diastolic and end-systolic volumes. Total end-diastolic and end-systolic cavity volumes were computed by the workstation using a modified Simpson’s rule equation. Stroke volume (SV) was calculated as the difference between end-diastolic and end-diastolic volumes, the product of SV and heart rate was defined as CO. 2.5 Statistical analysis Statistical comparison of EV and CMR values was performed as proposed by Bland and Altman to avoid the methodological problems introduced by correlation coefficients and regression analysis [20]. The arithmetic mean of the repeated EV measurements was compared to values determined by CMR. The percentage error was calculated from the limits of agreement and mean COCMR. For evaluation of the comparability we used the approach of Critchley and Critchley [21] whereupon two methods for measurement of CO can be considered equivalent if the limits of agreement do not exceed ±30 %. This would be equivalent to 1.5 l/min assuming a mean CO of 5 l/min and a mean bias of 0.2 l/min. With a corresponding standard deviation of 0.75 l/min we calculated that a planned sample size of 113 patients would give the study a power of 80 % at a 5 % significance level to detect a difference between the two methods. Student’s t test, Mann–Whitney test and Fisher’s exact test were used for univariate analysis. Analyses of variances (ANOVA) and multiple regression analysis were applied in multivariate testing. Pearson product-moment correlation coefficient was used. An alpha error of \5 % was considered statistically significant.

2.4 Cardiac magnetic resonance imaging

3 Results Determination of CO by CMR was performed on a 1.5 T whole-body imaging system (Magnetom Sonata or Avanto, Siemens Medical Systems, Erlangen, Germany). Images were acquired during repeated end-expiratory breath-holds starting with scout images. ECG-gated cine images were then acquired using a segmented steady state free precession sequence (TrueFISP) resulting in three long-axis views and 7–12 short-axis views averaged over multiple heart cycles depending on heart rate. End-diastole and endsystole were defined as the frames revealing the largest and smallest cavity area, respectively. Epicardial and endocardial contours were outlined manually once on each enddiastolic and end-systolic short-axis frame. Papillary

A total of 18 patients (12 %) had to be excluded from final data analysis due to invalid EV measurements resulting in a collective of 134 patients. There was a significant difference for age and number of pack years as well as a trend for HREV between successful and unsuccessful measurements, respectively. Baseline characteristics are given in Table 2. CO ranged from 2.6 to 10.4 l/min with a median of 5.5 l/min as determined by CMR, and from 2.5 to 8.9 l/ min with a median of 4.4 l/min with EV. A summary of the hemodynamic parameters is shown in Table 3. There was significant difference between CMR and EV for measurements of CO (p \ 0.0001), SV (p \ 0.0001) and HR

123

J Clin Monit Comput

(p = 0.03) each. In Bland–Altman analysis we found insufficient agreement between the two methods with a mean bias of 1.2 ± 1.4 l/min or 23 ± 26 % (mean ± SD; Fig. 1a). The analysis of repeated measurements demonstrated a high reproducibility for EV with a mean bias of 0.0 ± 0.3 l/min or 0 ± 8 % (Fig. 1b). The percentage errors were 51 and 15 % for agreement and reproducibility, respectively. For the evaluation of factors that may influence the accuracy of EV results, an outlier analysis was performed comparing pairs of measurements with better or worse agreement, respectively. Outliers were defined as pairs of measurements with a difference [1.7 l/min between COCMR and COEV corresponding to 30 % of the mean COCMR. In univariate analysis we found a statistically significant difference between the two groups for gender, height, weight, CO and stroke volume determined by CMR (Table 4). In the stepwise multiple regression model gender, height, COCMR and SVCMR remained as independent predictors for a larger variation. Pearson correlation coefficients were 0.23 (p \ 0.01) between bias and height and 0.7 (p \ 0.0001) between bias and COCMR. When investigating the influence on reproducibility none of the five significant factors associated with larger variation remained in the stepwise multiple regression model. In general, we found a significantly larger bias of 1.5 ± 1.5 l/min for men as compared to 0.7 ± 1.2 l/min in women, respectively (p \ 0.01). We therefore performed further analyses for gender specific differences. Correlation of bias and COCMR was equal (p = 0.74) between men (r = 0.69, p \ 0.0001) and women (r = 0.72, p \ 0.0001). There was a trend towards a closer correlation of bias and height in female (r = 0.21, p = 0.13) as compared to male (r = 0.03, p = 0.76) patients not reaching statistical significance (p = 0.3). Accordingly, women also showed a

Table 2 Baseline characteristics

Variable

Table 3 Hemodynamic parameters (n = 134) Variable

Value

Range

COCMR (l/min)

5.5

2.6–10.4

SVCMR (ml)

82

38–175

HRCMR (min-1)

67

45–120 2.5–8.9

COEV (l/min)

4.4

SVEV (ml)

69

44–108

HREV (min-1)

65 ± 12

43–98

BPsystolic (mmHg)

133

96–207

BPdiastolic (mmHg)

82

58–142

Values not distributed normally are presented as median, normally distributed data as mean ± SD CO cardiac output, SV stroke volume, HR heart rate, BP blood pressure, CMR cardiac magnetic resonance imaging, EV electrical velocimetry

trend (p = 0.10) for closer correlation of bias and weight (r = 0.37, p \ 0.01) when being compared to men (r = 0.09, p = 0.43). In contrast, no significant difference could be found for reproducibility between men (0.0 ± 0.3 l/min) and women (0.0 ± 0.4 l/min), respectively (p = 0.94). Since extreme CO states are known to be associated with inaccuracies for several techniques, measurement bias was stratified in quintiles of COCMR for further evaluation. In the first quintile including CO values between 2.6 and 4.4 l/min we found a good agreement of -0.2 ± 0.9 l/min (-2 ± 22 %). However, there was a significantly larger bias with increasing quintiles reaching values as high as 3.1 ± 1.1 l/min (48 ± 20 %) in the fifth one (p \ 0.001, ANOVA; Fig. 2). In contrast, reproducibility was not negatively affected by mean COEV quintiles (p = 0.71, ANOVA).

Overall (n = 152) Value

Range

Age (years)

54

Male gender

95 (63 %)

15–86

Height (cm)

173 ± 9

Weight (kg)

82

50–140

HREV (min-1)

66 ± 13

42–99

BMI (kg/m2)

27.1 ± 4.6

17.7–40.5

Smoker

69 (45 %)

Pack years

16

EV successful (n = 134)

EV unsuccessful (n = 18)

Value

Value

53

Range 15–83

83 (62 %) 150–199

173 ± 9

Range 35–86

0.03*

12 (67 %)

0.80

150–199

174 ± 7

160–190

0.73

81

50–140

84 ± 14

54–103

0.25

65 ± 12

43–98

71 ± 17

42–99

0.10

27 ± 4.6

17.7–40.5

18.7–34

0.44

58 (43 %) 1–140

62 ± 14

p value

15

27.9 ± 4.5 11 (61 %)

1–100

52 ± 48

0.21 1–140

0.03*

Values not distributed normally are presented as median, normally distributed data as mean ± SD HREV heart rate using electrical velocimetry, BMI body mass index * p \ 0.05

123

J Clin Monit Comput Fig. 1 Overall agreement and reproducibility. a Bland– Altman plot with insufficient agreement between CMR and EV with a mean bias of 1.2 ± 1.4 l/min (23 ± 26 %) and a percentage error of 51 %. b Bland–Altman plot showing a high reproducibility of 0.0 ± 0.3 l/min (0 ± 8 %, percentage error 15 %) for EV

Subgroup analyses were performed in patients with arrhythmias or changes in thoracic fluid content. A total of 13 patients with either atrial fibrillation or more than five ventricular or supraventricular extrasystoles at the time of EV measurement could be identified. They were pairmatched with patients in sinus rhythm for the parameters age, gender, height, weight, COCMR and HRCMR. Bland– Altman analysis showed a mean bias of 1.1 ± 0.8 l/min in the arrhythmia group as compared to 1.6 ± 0.8 l/min in

patients with sinus rhythm (p = 0.19). In 22 patients with pleural or pericardial effusion we found a mean bias of 0.9 ± 1.8 l/min versus 1.2 ± 1.3 l/min in pair-matched controls (p = 0.51). Moreover, pulmonary congestion due to left heart failure may lead to inaccuracies due to fluid imbalances. When comparing 67 patients with an ejection fraction \55 % in CMR to C55 % there was no significant difference with the mean bias being 1.1 ± 1.4 and 1.3 ± 1.5 l/min (p = 0.47), respectively.

123

J Clin Monit Comput Table 4 Outlier analysis

Variable

Non-outlier (n = 87)

Outlier (n = 47)

Value

Value

Range 15–82

51 ± 17

p value Range

Age (years)

52 ± 18

Male gender

47 (54 %)

17–83

Height (cm)

172 ± 9

150–199

176 ± 10

158–197

0.01*

Weight (kg)

78 ± 15

50–140

88 ± 18

55–140

\0.001* \0.0001*

36 (77 %)

0.60 0.01*

COCMR (l/min)

5.1 ± 1.1

2.6–8.3

6.8 ± 1.5

4.4–10.4

COEV (l/min)

4.4

2.5–7.4

4.0

2.6–8.9

SVCMR (ml)

75 ± 16

38–119

102 ± 25

60–175

SVEV (ml)

69 ± 13

44–108

69 ± 15

44–103

0.79

HRCMR (min-1)

69 ± 13

45–105

69 ± 16

45–120

0.88

HREV (min-1)

66 ± 11

44–93

63 ± 13

43–98

0.24

BPsystolic (mmHg) BPdiastolic (mmHg)

135 ± 21 82

96–182 61–112

137 ± 23 82

104–207 58–142

0.56 0.40

0.12 \0.0001*

Values not distributed normally are presented as median, normally distributed data as mean ± SD CO cardiac output, SV stroke volume, HR heart rate, BP blood pressure * p \ 0.05

Fig. 2 Circulatory condition classes. Bias between CMR and EV measurements separated in quintiles. Quintile 1 (2.6–4.4 l/ min): -0.2 ± 0.9 l/min (-2 ± 22 %), quintile 2 (4.5–5.2 l/min): 0.9 ± 0.6 l/min (23 ± 16 %), quintile 3 (5.3–5.8 l/min): 1.0 ± 1.2 l/min (22 ± 24 %), quintile 4 (5.9–6.9 l/min): 1.4 ± 1.0 l/min (26 ± 21 %), quintile 5 (7.0–10.4 l/min): 3.1 ± 1.1 l/ min (48 ± 20 %). Connectors indicate significant difference (p \ 0.05, ANOVA)

4 Discussion The aim of our study was to evaluate EV as a novel algorithm of impedance cardiography for the non-invasive determination of CO. In contrast to previous studies using mostly thermodilution as reference technique, values obtained by CMR as the non-invasive gold standard were used for comparison. Overall, there was an insufficient agreement between the two methods whereas reproducibility was high. Our large and heterogeneous collective

123

allowed performing detailed additional analyses to reveal potential sources of error. We performed univariate outlier analysis being defined as a 30 % bias according to current recommendations [21]. There were statistically significant differences between the two groups for gender, height, weight, CO and stroke volume determined by CMR. Except for weight, any of these remained as independent predictors for a larger variation in the multiple regression model. Since VETP used for absolute calculation of CO is estimated from body height and mass, these findings may

J Clin Monit Comput

correspond to inaccuracies in the underlying algorithm given the fact that there was significantly larger bias in men. This assumption is supported by the findings of Woltjer et al. investigating two algorithms of ICG in obese and normal weight patients undergoing bypass surgery. They could demonstrate that the weight correction implemented in the Sramek and Bernstein approach significantly influences the accuracy of stroke volume determination. In contrast, Kubicek’s method was interestingly not biased by weight [22]. While correlation of bias and COCMR was not different between genders, height and weight in our study each showed a trend to a closer correlation in women. On the one hand, this could be evidence for stronger factors influencing accuracy in men not targeted in this study. On the other hand, height, weight as well as greater variance in body composition and cross sectional area of the thorax may explain variation to some extent in women. Another explanation could be electrode arrangement which was demonstrated to have considerable impact on individual impedance measurements in healthy volunteers [23]. Small yet measureable changes could also be detected when changing electrode position in a group of surgical patients using another algorithm referred to as Physio FlowÒ [24]. In our study we used a standard inter-electrode gap of 5 cm according to manufacturer recommendation with additional research enduring for EV. When assessing reproducibility no influence of gender, height, weight, CO or SV as determined by CMR could be found. Consequently, intraindividual changes of CO seem to be robust to invariant factors such as gender or height. In contrast, absolute values as well as measurements taken after extreme changes in weight should be interpreted with caution. Extreme CO states are a known impediment to most measurement techniques. We found a good agreement in patients with hypodynamic states of -0.2 ± 0.9 l/min (-2 ± 22 %) and an increasing bias significantly underestimating CMR up to 3.1 ± 1.1 l/min (48 ± 20 %) in patients with increased CO. For thermodilution itself being the predominant reference technique an increased bias could be detected in patients with both high ([7 l/min) and low (\4 l/min) CO values [25–27]. For the novel non-invasive techniques a significantly reduced accuracy in extreme circulatory conditions is common when being compared to CMR. High CO values are underestimated using IGR ([6.3 l/min [28]), pulse contour analysis (PCA, [6.3 l/min [29]) and continuous-wave Doppler (CWD,[5.4 l/min [30, 31]). Additionally, low values were overestimated in IGR (\4.1 l/min) and PCA (\4.8 l/min). CWD showed a trend to a worse accuracy in patients with high BMI which however did not reach statistical significance. Reproducibility was not negatively affected by high or low CO states which could also be demonstrated for IGR, PCA and

CWD. However, PCA showed a significantly worse reproducibility in outliers [29]. We performed subgroup analyses in patients with rhythm disorders [3, 4, 32] and changes in thoracic fluid content [33–35] being both known impediments to impedance cardiography. In a subgroup of 13 patients with arrhythmia due to either atrial fibrillation or more than five ventricular or supraventricular extrasystoles no difference in measurement bias was found. We used averaged CO values collected over an interval of 20 s in order to compensate for respiratory or arrhythmic hemodynamic changes possibly explaining the missing impairment. This is similar to IGR which was accordingly demonstrated to be robust against rhythm disorders [36]. Notably, 12 % of our measurements had to be excluded from final data analysis. Although this is comparable to IGR [28], smaller numbers were postulated for a technique not depending on patient collaboration such as EV. Besides higher heart rates leading e.g. to an impaired detection of the ECG signal, large numbers of pack years and age could be identified as factors associated with a higher probability of unsuccessful EV measurements. For ICG it could be demonstrated that changes in thoracic fluid content caused by pulmonary edema or pleural effusion negatively affected measurement accuracy [34]. The technique is even proposed for monitoring fluid changes postulating a possible use in the early detection of intrapulmonary fluid accumulation [35, 37]. Although not being primarily based on the changes of absolute thoracic blood volume like impedance cardiography, EV uses aortic blood flow velocity and ejection time for the calculation of CO. Nevertheless, base impedance as well as thoracic fluid content derived from body height and mass are incorporated in the model. In our investigation, neither patients in a subgroup with pleural or pericardial effusion nor in a subgroup with an ejection fraction \55 % did show a significantly increased bias. Relevant pleural or pericardial effusions as well as congestion yet are likely to be underrepresented in our CMR collective. Despite the insufficient agreement with the non-invasive gold standard, its high reproducibility may be an important advantage of EV in serial measurements while no information on tracking intra-individual changes of CO can be drawn from our data. Previous studies proposed several fields of application in stable patients: Voss et al. used the technique to systematically evaluate the impact of basic pacing rate on hemodynamic parameters in cardiac resynchronization therapy in contrast to the routinely performed echocardiographic atrioventricular interval optimization. They could demonstrate that for heart rates between 40 and 70 min-1, an increase in basic pacing rate enhances CO without reducing stroke volume [38]. Using

123

J Clin Monit Comput

a different algorithm, ICG was found to be feasible for the hemodynamic assessment during 6-min walk test in patients suffering from pulmonary hypertension [39]. Similarly, hemodynamic measurements using IGR were identified as potential alternative outcome measures in this collective. Besides being used as a measure for treatment response hemodynamic parameters appear to be more sensitive to change than 6-min walk distance in fitter patients [40]. Flinck et al. could demonstrate a significant negative correlation between aortic attenuation as a measure of coronary artery attenuation and CO determined by EV in patients undergoing computed tomography coronary angiography. The resulting formula made it possible to calculate required contrast agent volumes according to the circulatory condition without lowering diagnostic quality [41]. Other possible fields of application in clinically stable patients could e.g. include differential diagnosis of syncope or monitoring of cardiotoxic chemotherapy. Despite the low operating costs and non-invasiveness, further investigation is required prior to any of the stated applications in clinical routine. Especially trending should be focused considering the insufficient agreement found in our study. Although hemodynamic assessment may be additionally valuable under unstable circulation, available studies yielded diverse results. When comparing EV to transcardiopulmonary thermodilution in septic patients after major general surgery only poor agreement was found for CO [12]. In contrast, Zoremba et al. [11] found a good agreement between EV and transcardiopulmonary and pulmonary arterial thermodilution each. Recently, it could be demonstrated that EV enables clinicians to determine hemodynamic changes prior to onset of hypotension in patients under spinal anesthesia for cesarean delivery [42]. For a different algorithm referred to as bioreactance Kupersztych–Hagege et al. also found insufficient agreement when being compared to thermodilution in a medical intensive care unit. Additionally, the technique was unable to predict fluid responsiveness through the passive leg raise test or track changes in CO following volume expansion [43]. Several factors may contribute to larger variations in critically ill patients. In sepsis an appreciable portion of continuous rather than pulsatile flow may lead to inaccuracies when the pulsatile component during systole is considered like in ICG [3]. Known impediments such as changes in thoracic fluid content or arrhythmias are more likely to be present. Moreover, high CO states seen in e.g. septic shock were shown to be associated with worse accuracy even under stable conditions in our investigation. Since reproducibility can be further impaired by hemodynamic instability, additional investigation is also needed in critically ill patients.

123

5 Limitations When interpreting the present results, several points should be taken into consideration. First, only patients with indication for CMR were included leading to a possibly selection bias as instable patients are underrepresented. Second, the simultaneous measurement of hemodynamic parameters is desirable to exclude hemodynamic changes between the measurements. Because EV is not suitable for CMR imaging, the measurements had to be taken either directly before or after the CMR examination. However, because both measurements were taken after stabilization of the heart rate and blood pressure, equality of circulatory conditions can be assumed. Third, no information on trending can be drawn from our investigation. Fourth, stroke volume varies with respiratory cycle due to changes in left ventricular filling with the negative or positive intra-thoracic pressure. While CMR measurements were taken during end-expiratory breath-holds, EV calculations are based on averaged values over the complete respiratory cycle therefore affecting the comparability of both methods.

6 Conclusion Electrical velocimetry is feasible for the determination of CO. It is associated with low operating costs requiring only few expendable items. However, overall agreement with cardiac magnetic resonance imaging is insufficient showing an increased bias with gender, height and increasing CO as well as stroke volume in stable patients. Arrhythmias or changes in thoracic fluid content had no effect on either accuracy or reproducibility. Overall short-term reproducibility was high and not negatively affected by factors influencing measurement bias. Therefore, absolute values acquired by EV should not be used interchangeably in clinical routine. EV may find its place for clinical application in intensive or intermediate care as well as stable patients, yet further investigation on its trending ability in different clinical situations is still pending. Compliances with Ethical Standards Conflict of interest of interest.

The authors declare that they have no conflict

Ethical statement The study protocol was approved by the institutional Ethics Committee and conducted in accordance with the Declaration of Helsinki. Informed consent was obtained from all participants.

References 1. Pohost GM, Hung L, Doyle M. Clinical use of cardiovascular magnetic resonance. Circulation. 2003;108(6):647–53.

J Clin Monit Comput 2. Kubicek WG, Karnegis JN, Patterson RP, Witsoe DA, Mattson RH. Development and evaluation of an impedance cardiac output system. Aerosp Med. 1966;37(12):1208–12. 3. Appel PL, Kram HB, Mackabee J, Fleming AW, Shoemaker WC. Comparison of measurements of cardiac output by bioimpedance and thermodilution in severely ill surgical patients. Crit Care Med. 1986;14(11):933–5. 4. Franko E, Van De Water J, Wang X. Ideal measurement of cardiac output: is impedance cardiography the answer? Vasc Endovascular Surg. 1991;25:550–8. 5. Carvalho P, Paiva R, Henriques J, Antunes M, Quintal I, Muehlsteff J. Robust characteristic points for ICG-definition and comparative analysis. Biosignals Rome, Italy; 2011. p. 161–8. 6. Norozi K, Beck C, Osthaus WA, Wille I, Wessel A, Bertram H. Electrical velocimetry for measuring cardiac output in children with congenital heart disease. Br J Anaesth. 2008;100(1):88–94. 7. Osthaus WA, Huber D, Beck C, Winterhalter M, Boethig D, Wessel A, Sumpelmann R. Comparison of electrical velocimetry and transpulmonary thermodilution for measuring cardiac output in piglets. Paediatr Anaesth. 2007;17(8):749–55. 8. Schmidt C, Theilmeier G, Van Aken H, Korsmeier P, Wirtz SP, Berendes E, Hoffmeier A, Meissner A. Comparison of electrical velocimetry and transoesophageal Doppler echocardiography for measuring stroke volume and cardiac output. Br J Anaesth. 2005;95(5):603–10. 9. Suttner S, Schollhorn T, Boldt J, Mayer J, Rohm KD, Lang K, Piper SN. Noninvasive assessment of cardiac output using thoracic electrical bioimpedance in hemodynamically stable and unstable patients after cardiac surgery: a comparison with pulmonary artery thermodilution. Intensive Care Med. 2006;32(12):2053–8. 10. Tomaske M, Knirsch W, Kretschmar O, Woitzek K, Balmer C, Schmitz A, Bauersfeld U, Weiss M. Cardiac output measurement in children: comparison of Aesculon cardiac output monitor and thermodilution. Br J Anaesth. 2008;100(4):517–20. 11. Zoremba N, Bickenbach J, Krauss B, Rossaint R, Kuhlen R, Schalte G. Comparison of electrical velocimetry and thermodilution techniques for the measurement of cardiac output. Acta Anaesthesiol Scand. 2007;51(10):1314–9. 12. Raue W, Swierzy M, Koplin G, Schwenk W. Comparison of electrical velocimetry and transthoracic thermodilution technique for cardiac output assessment in critically ill patients. Eur J Anaesthesiol. 2009;26(12):1067–71. 13. Petter H, Erik A, Bjorn E, Goran R. Measurement of cardiac output with non-invasive Aesculon impedance versus thermodilution. Clin Physiol Funct Imaging. 2011;31(1):39–47. 14. Heringlake M, Handke U, Hanke T, Eberhardt F, Schumacher J, Gehring H, Heinze H. Lack of agreement between thermodilution and electrical velocimetry cardiac output measurements. Intensive Care Med. 2007;33(12):2168–72. 15. Tomaske M, Knirsch W, Kretschmar O, Balmer C, Woitzek K, Schmitz A, Bauersfeld U, Weiss M. Evaluation of the Aesculon cardiac output monitor by subxiphoidal Doppler flow measurement in children with congenital heart defects. Eur J Anaesthesiol. 2009;26(5):412–5. 16. Trinkmann F, Berger M, Hoffmann U, Borggrefe M, Kaden JJ, Saur J. A comparative evaluation of electrical velocimetry and inert gas rebreathing for the non-invasive assessment of cardiac output. Clin Res Cardiol. 2011;100(10):935–43. 17. Grollmuss O, Demontoux S, Capderou A, Serraf A, Belli E. Electrical velocimetry as a tool for measuring cardiac output in small infants after heart surgery. Intensive Care Med. 2012; 38(6):1032–9. 18. Bernstein D, Osypka M. Apparatus and method for determining an approximation of the stroke volume and the cardiac output of the heart. US Patent 6,511,438 B2; 2003.

19. Papavassiliu T, Kuhl HP, Schroder M, Suselbeck T, Bondarenko O, Bohm CK, Beek A, Hofman MM, van Rossum AC. Effect of endocardial trabeculae on left ventricular measurements and measurement reproducibility at cardiovascular MR imaging. Radiology. 2005;236(1):57–64. 20. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1(8476):307–10. 21. Critchley LA, Critchley JA. A meta-analysis of studies using bias and precision statistics to compare cardiac output measurement techniques. J Clin Monit Comput. 1999;15(2):85–91. 22. Woltjer HH, Bogaard HJ, van der Spoel HI, de Vries PM. The influence of weight on stroke volume determination by means of impedance cardiography in cardiac surgery patients. Intensive Care Med. 1996;22(8):766–71. 23. Hoetink AE, Faes TJ, Schuur EH, Gorkink R, Goovaerts HG, Meijer JH, Heethaar RM. Comparing spot electrode arrangements for electric impedance cardiography. Physiol Meas. 2002;23(2): 457–67. 24. Tan KH, Lai FO, Hwang NC. Measurement of cardiac output using Physio Flow with different positions of electrode placement. Singapore Med J. 2006;47(11):967–70. 25. Dhingra VK, Fenwick JC, Walley KR, Chittock DR, Ronco JJ. Lack of agreement between thermodilution and fick cardiac output in critically ill patients. Chest. 2002;122(3):990–7. 26. van Grondelle A, Ditchey RV, Groves BM, Wagner WW Jr, Reeves JT. Thermodilution method overestimates low cardiac output in humans. Am J Physiol. 1983;245(4):H690–2. 27. Stetz CW, Miller RG, Kelly GE, Raffin TA. Reliability of the thermodilution method in the determination of cardiac output in clinical practice. Am Rev Respir Dis. 1982;126(6):1001–4. 28. Saur J, Fluechter S, Trinkmann F, Papavassiliu T, Schoenberg S, Weissmann J, Haghi D, Borggrefe M, Kaden JJ. Noninvasive determination of cardiac output by the inert-gas-rebreathing method—comparison with cardiovascular magnetic resonance imaging. Cardiology. 2009;114(4):247–54. 29. Trinkmann F, Sampels M, Doesch C, Papavassiliu T, Brade J, Schmid-Bindert G, Hoffmann U, Borggrefe M, Kaden JJ, Saur J. Is arterial pulse contour analysis using Nexfin a new option in the noninvasive measurement of cardiac output?—A pilot study. J Cardiothorac Vasc Anesth. 2013;27(2):283–7. 30. Trinkmann F, Doesch C, Papavassiliu T, Weissmann J, Haghi D, Gruettner J, Schoenberg SO, Borggrefe M, Kaden JJ, Saur J. A novel noninvasive ultrasonic cardiac output monitor: comparison with cardiac magnetic resonance. Clin Cardiol. 2010;33(2):E8–14. 31. Tan HL, Pinder M, Parsons R, Roberts B, van Heerden PV. Clinical evaluation of USCOM ultrasonic cardiac output monitor in cardiac surgical patients in intensive care unit. Br J Anaesth. 2005;94(3):287–91. 32. Williams BO, Caird FI. Accuracy of the impedance cardiogram in the measurement of cardiac output in the elderly. Age Ageing. 1985;14(5):277–81. 33. Berman IR, Scheetz WL, Jenkins EB, Hufnagel HV. Transthoracic electrical impedance s a guide to intravascular overload. Arch Surg. 1971;102(1):61–4. 34. Penney BC. Theory and cardiac applications of electrical impedance measurements. Crit Rev Biomed Eng. 1986;13(3):227–81. 35. Pomerantz M, Delgado F, Eiseman B. Clinical evaluation of transthoracic electrical impedance as a guide to intrathoracic fluid volumes. Ann Surg. 1970;171(5):686–94. 36. Saur J, Trinkmann F, Doesch C, Weissmann J, Hamm K, Schoenberg SO, Borggrefe M, Haghi D, Kaden JJ. Non-invasive measurement of cardiac output during atrial fibrillation: comparison between cardiac magnetic resonance imaging and inert gas rebreathing. Cardiology. 2010;115(3):212–6.

123

J Clin Monit Comput 37. Tang WH, Tong W. Measuring impedance in congestive heart failure: current options and clinical applications. Am Heart J. 2009;157(3):402–11. 38. Voss F, Becker R, Hauck M, Katus HA, Bauer A. The basic pacing rate in CRT patients: the higher the better? Clin Res Cardiol. 2009;98(4):219–23. 39. Tonelli AR, Alkukhun L, Arelli V, Ramos J, Newman J, McCarthy K, Pichurko B, Minai OA, Dweik RA. Value of impedance cardiography during 6-minute walk test in pulmonary hypertension. Clin Transl Sci. 2013;6(6):474–80. 40. Lee WT, Brown A, Peacock AJ, Johnson MK. Use of non-invasive haemodynamic measurements to detect treatment response in precapillary pulmonary hypertension. Thorax. 2011;66(9):810–4.

123

41. Flinck M, Graden A, Milde H, Flinck A, Hellstrom M, Bjork J, Nyman U. Cardiac output measured by electrical velocimetry in the CT suite correlates with coronary artery enhancement: a feasibility study. Acta Radiol. 2010;51(8):895–902. 42. Liu Y, Pian-Smith MC, Leffert LR, Minehart RD, Torri A, Cote C, Kacmarek RM, Jiang Y. Continuous measurement of cardiac output with the electrical velocimetry method in patients under spinal anesthesia for cesarean delivery. J Clin Monit Comput. 2014. doi:10.1007/s10877-014-9645-8. 43. Kupersztych-Hagege E, Teboul JL, Artigas A, Talbot A, Sabatier C, Richard C, Monnet X. Bioreactance is not reliable for estimating cardiac output and the effects of passive leg raising in critically ill patients. Br J Anaesth. 2013;111(6):961–6.