a b s t r a c t

Background: To determine the optimum chest compression site during cardiopulmonary resuscitation (CPR) with regard to heart failure by applying three-dimensional (3D) coordinates on computed tomography (CT). Methods: This retrospective, cross-sectional study involved adults who underwent echocardiography and CT on the same day from 2007 to 2017. Incomplete CT images or information on HF, cardiac medication between echo- cardiography and CT, or thoracic abnormalities were excluded. Cases were checked whether they had HF through symptom/sign assessment, N-terminal pro-B type natriuretic peptide, and echocardiography. We set the xiphisternal joint’s midpoint as the reference (0, 0, 0) to draw a 3D coordinate system, designating leftward, up- ward, and into-the-thorax directions as positive. The coordinate of the maximum LV diameter’s midpoint (P_max.LV) was identified.

Results: Enrolled were 148 patients (63.0 +- 15.1 years) with 87 females and 76 HF cases. P_max.LV of HF cases was located more leftwards, lower, and deeper than non-HF cases (5.69 +- 0.98, -1.51 +- 1.67, 5.76 +- 1.09 cm

vs. 5.00 +- 0.83, -0.99 +- 1.36, 5.25 +- 0.71 cm, all p b 0.05). Fewer HF cases had their LV compressed than

non-HF cases (59.2% vs. 77.8%, p = 0.025) when being compressed according to the current guidelines. The aorta (vs. LV) was compressed in 85.5% and 81.9% of HF and non-HF cases, respectively, at 3 cm above the xiphisternal joint. At 6 cm above the joint, the highest allowable position according to the current guidelines, all victims would have their aorta compressed directly during CPR rather than the LV. Conclusions: The lowest possible sternum just above the xiphisternal joint should be compressed especially for HF patients during CPR.

(C) 2017

Introduction

Half of the out-of-hospital cardiac arrest patients have known cardiac disease, and one-fourth of them have overt heart failure [1]. Sudden cardiac death is the primary (~ 42%) and secondary (26-28%) cause of death in HF patients with reduced and preserved ejec- tion fraction (EF), respectively [2-4]. Hence, effective cardiopulmonary resuscitation (CPR) is critical to save HF patients in CA.

In 2010, the airway-breathing-circulation approach was changed to circulation-airway-breathing approach, which highlights the impor- tance of chest compression in CPR [5,6]. High-quality CPR is achieved by ensuring adequate rate and depth of chest compression, allowing

? This study has been presented at the 30th European Society of Intensive Care Medicine (ESICM) Annual Congress, which was held in Vienna from 23th to 27th Sep, 2017.

?? This study was not funded at all.

* Corresponding author at: Department of Emergency Medicine, CHA Bundang Medical Center, CHA University, 59, Yatap-ro, Bundang-gu, Seongnam-si, 13496 Gyeonggi-do, Republic of Korea.

E-mail address: [email protected] (S.-B. Chon).

full chest recoil between compressions, minimizing interruptions in chest compression, and avoiding excessive ventilation to re-emphasize effective chest compression [6,7]. Both the European Resuscitation Coun- cil and American Heart Association still recommend compression of the ‘lower half’ of the sternum [5,6], although some researchers suggested that the ‘lowest’ part should be compressed to maximize the stroke vol- ume by compressing the maximum diameter of the Left ventricle and avoiding compression of the aorta [8-10]. The rationale for the cur- rent, unchanged hand positioning recommendation may probably be the fear of incurring Possible complications, such as liver, stomach, or spleen injury [11,12].

HF is usually accompanied by cardiomegaly [13,14]. Considering that

the heart is influenced by gravity and hung on the relatively fixed aorta and superior and inferior vena cava, an enlarged heart may be shifted in- feriorly and leftwards. It implies that the optimum chest compression site for HF patients may also move towards the same direction, challeng- ing again the currently recommended universal chest compression site. Our primary aim was to determine the optimum chest compression site by comparing the patients with and without HF. We introduced a three-dimensional (3D) coordinate on chest computed tomography

https://doi.org/10.1016/j.ajem.2017.07.041

0735-6757/(C) 2017

(CT) to obtain the accurate spatial information of intrathoracic struc- tures. To check the compatibility of these optimum compression sites with the current CPR guidelines was the secondary goal.

Materials and methods

Setting and participants

A retrospective, cross-sectional study was performed in a N 800-bed university hospital in Korea.

All adults aged >=18 years who underwent transthoracic echocardiog-

raphy and CT on the same day from Mar 2007 to Feb 2017 were consid- ered eligible. The exclusion criteria are as follows: (1) heart and thorax were not delineated entirely on CT images; (2) administration of crystalloid-loading, Antihypertensive drugs including diuretics, inotropic or chronotropic agents b 2 h before or between performing echocardiog- raphy and CT, which could prohibit the presumption of simultaneity for these two studies by affecting the cardiac function; (3) emergency de- partment cases, most of which needed Urgent treatment, resulting in the condition stated in (2); (4) thoracic abnormalities, which could affect the location of the heart, such as N 5 mm-depth pericardial effusion, peri- cardial tumour/cyst, atelectasis, lobectomy, N 10% pneumothorax, N 2 cm- depth unilateral pleural effusion, diaphragmatic hernia, and funnel chest

[14]; and (5) unclear information on the presence/absence of HF.

All cases were classified based on the presence/absence of HF. HF cases were defined as (1) LV with b 40% EF on echocardiography per- formed by cardiologists (HF with reduced EF [HFrEF]), or (2) having the following 3 criteria altogether, otherwise known as HF with pre- served EF (HFpEF): (i) typical HF symptoms/signs, (ii) N 125 pg/mL N- terminal pro-B type natriuretic peptide (NT-proBNP), and (iii) relevant structural heart disease (>= 115 and >= 95 g/m² LV mass index for men and

women, respectively, or N 34 mL/m² left atrial volume index) or diastolic

dysfunction (>= 13 mitral peak velocity of early filling/early diastolic mi- tral annulus velocity (E/e?)) [15,16]. patients without HF were defined as (1) >= 50% EF, and (2) <= 125 pg/mL NT-proBNP or absence of echocar- diographic evidences of structural heart disease or diastolic dysfunction

[15]. To eliminate bias, information on the presence/absence of HF was concealed while collecting other information.

Data collection and measurement

For the demographic characteristics, data on the sex, age, height, and weight were obtained to calculate the body mass index (BMI = height / weight²). Information on comorbidity regarding hypertension; diabetes mellitus; hyperlipidaemia; stroke; coronary artery disease; chronic kid- ney, liver, and lung diseases; tuberculosis; and malignancy were collect- ed. The main department that managed the patients was identified.

The lengths of the whole sternum and xiphoid process were mea-

sured on CT images. Subsequently, the midpoint of the lower half of the sternum was identified as the representative guideline point (P_guideline) as the current CPR recommendations advise to compress the lower half of the sternum (Fig. 1) [5,6,8]. The left cardiac structure just beneath the P_guideline, which would be compressed directly dur- ing CPR based on the current guidelines, was also identified as either the LV or aorta (within or beyond the root of aorta).

A 3D coordinate system was designed to delineate the exact spatial information on the intrathoracic structures (Fig. 2). The reference point (P_reference) with its own horizontal (x), vertical (y), and deep

(z) axes of coordinate (0, 0, 0) was set at the midpoint of the xiphisternal joint, where the sternal body, xiphoid process, and both costal margins meet. From the P_reference, the leftward, upward, and into-the-thorax directions were designated as positive on the x, y, and z axes, respective- ly, all of which form right angles to one another.

Using the CT’s navigating function, the sagittal image that showed the maximum LV diameter was chosen (Fig. 3A). The 3D coordinates of the midpoint (P_max.LV), which bisect this maximum diameter,

Fig. 1. Measurement of the lengths of the whole sternum and xiphoid process. The lengths of the sternum and xiphoid process were measured on the mid-sagittal image using the gauging function of the CT. (P_guideline, the midpoint of the lower half of the sternum representing the chest compression site based on the current European Resuscitation Council and American Heart Association recommendations; P_reference, the reference point with its own (0, 0, 0) coordinate set at the midpoint of the xiphisternal joint, where the sternal body, xiphoid process, and both costal margins meet.)

was identified using the measuring function directly or multiplying the predetermined thickness (2, 3, or 5 mm) per slice by the number of in-between slices as (x_max.LV, y_max.LV, z_max.LV) (Fig. 3D and

Fig. 2. A three-dimensional (3D) coordinate system used to delineate the exact spatial information of intrathoracic structures. The reference point (P_reference) with its own (0, 0, 0) coordinate was set at the midpoint of the xiphisternal joint, where the sternal body, xiphoid process, and both costal margins meet. From the P_reference, the leftward, upward, and into-the-thorax directions were designated as positive on the horizontal (x), vertical (y), and deep (z) axes, respectively, all of which made right angles to one another. Using the 3D coordinates, the distance between P₁ (x₁, y₁, z₁) and

P₂ (x₂, y₂, z₂) can be calculated as ?(x₁ - x₂)² + (y₁ - y₂)² + (z₁ - z₂)², based on the

Pythagorean theorem.

Fig. 3. Identification of the 3D coordinates of (1) the midpoint of the maximum left ventricle diameter (P_max.LV) and (2) the centre point of the aortic annulus (P_aorta). (1) Using the navigating function of CT, the sagittal image that showed the maximum LV diameter was chosen (Panel A). The midpoint (P_max.LV) was identified to bisect this maximum diameter (Panel D, empty circle). To determine the coordinate of P_max.LV (x_max.LV, y_max.LV, z_max.LV), P_max.LV’ was defined as the crossing point where the vertical line projecting from the P_max.LV penetrates the midsagittal plane (x = 0), hence having its 3D coordinates (0, y_max.LV, z_max.LV) (Panel B, empty circle). Now, P_max.LV and P_max.LV’ (empty circles) share the same y and z coordinates, and these were measured through the gauging function of CT at the midsagittal plane (x = 0) (Panel B). Regarding x_max.LV, it was calculated by multiplying the number of CT slices by the predetermined thickness (2, 3, or 5 mm) (Panel A). (2) A similar approach was utilized to delineate the 3D coordinates (x_aorta, y_aorta, z_aorta) of the centre point of the aortic annulus (ventriculo-aortic junction, P_aorta) by defining the P_aorta’ (0, y_aorta, z_aorta) at the midsagittal plane again (Panels A, B, and C, black circles). (3D, three dimensional; P_reference, the reference point with its own (0, 0, 0) coordinate set at the midpoint of the xiphisternal joint (black rectangle), where the sternal body, xiphoid process, and both costal margins meet).

B). We assumed that the optimum compression site should be the clos- est one on the anterior chest wall to this P_max.LV [8-10]. Likewise, the 3D coordinates of the centre point of the aortic annulus (ventriculo-aor- tic junction, P_aorta) was identified as (x_aorta, y_aorta, z_aorta) (Fig. 3A, B, and C). We calculated the distance between the P_max.LV

and P_aorta, which is known to influence the LV stroke volume [9], through the Pythagorean theorem as follows: ?(x_max.LV - x_aorta)²

+ (y_max.LV - y_aorta)² + (z_max.LV - z_aorta)² (Fig. 2).

Statistical analysis

Type 1 and 2 errors were set as 0.05 and 0.10, respectively. The dis- tance difference on each axis was assumed to be 0.6 cm, which would re- sult in 1.0 (??(0.6)² + (0.6)² + (0.6)²) cm difference in 3D space, with standard deviation of 1.0 cm. Then, 58 cases per each group were

required.

The demographic, clinical, echocardiographic, and CT charac- teristics and NT-proBNP concentrations were shown based on the absence/presence of HF. These were compared using chi-square and t- test for categorical and continuous variables, respectively.

Subgroup analysis was performed similarly within HF group to com- pare the HFrEF and HFpEF cases.

As a sensitivity analysis to determine the effect of the varying height (y_sternum) of the hand positioning (P_sternum) on the sternum, the

height difference between P_aorta and P_sternum was calculated (Fig. 4). When P_aorta was more superior to P_sternum, compression of the P_sternum would result in compression of the LV rather than the

aorta. We set the y_sternum from -3 to +6 cm (9 cm), the full range of the lower half of the whole sternum (18 cm), with 1 cm differences.

All the statistical analyses were performed using IBM SPSS Statistics, version 24 (SPSS Inc., Chicago, IL, USA). Statistical significance was pre- sumed when p b 0.05. The data were presented as mean +- standard deviation.

The Institutional Review Board approved this study (no. 2017-04- 003-002). Informed consents were waived.

Results

A total of 148 patients, with mean age of 63.0 +- 15.1 years, were en- rolled (Fig. 5). Eighty-seven (58.8%) were female. Seventy-six (51.4%) cases had HF, with HFpEF and HFrEF accounting for 37 and 39 of the total cases, respectively. The NT-proBNP in 2 HFpEF and 4 HFrEF cases was N 35,000 pg/mL, the upper measurable limit. For statistical analysis, these values were replaced with 35,000 pg/mL.

HF patients were older and had more chronic kidney and lung dis- eases and higher NT-proBNP, smaller EF, and bigger E/e?, left atrium vol- ume, and LV mass indexes than non-HF patients (Table 1).

Fig. 4. A schematic diagram to perform a sensitivity analysis to calculate the effect of varying hand positioning heights on the sternum (P_sternum) on (1) the height difference between the sternum and P_aorta and (2) the proportion that the compression of the P_sternum is supposed to compress the LV rather than the aorta. The height of P_sternum ranged from -3 cm to +6 cm, and each height is shown with ‘x’ mark. P_aorta’ and P_max.LV’, which projected vertically from P_aorta and P_max.LV to the anterior thoracic surface (plane at z = o), are shown. Here, the mean values of the 3D coordinates of HF cases were used: (1.63, 1.16, 0 cm) and (5.69, -1.51, 0 cm), respectively. The 3D reconstructed skeleton image, which reconstructs the bony structures only, reveals that the P_max.LV’ is located on the cartilaginous portion (sternocostal joint) of the left lower anterior thorax. The results of the sensitivity analysis are demonstrated in Table 3. (3D, three dimensional; HF, heart failure; LV, left ventricle; P_aorta, the centre point of the aortic annulus; P_max.LV, the midpoint

used to bisect the maximum LV diameter chosen among sagittal images; P_reference, the reference point with its own (0, 0, 0) coordinate set at the midpoint of the xiphisternal joint, where the sternal body, xiphoid process, and both costal margins meet; P_sternum, the point where the compressing hand is in contact with the sternum).

CT revealed no significant differences in the lengths of the whole ster- num and xiphoid process between the non-HF and HF patients. Howev- er, HF patients had significantly smaller proportion of LV being located just beneath the P_guideline than non-HF patients (59.2% vs. 77.8%, p

= 0.025). The P_max.LV of HF patients was located more leftwards, lower, and deeper than that of non-HF patients (5.69 +- 0.98, -1.51 +- 1.67, 5.76 +- 1.09 cm vs. 5.00 +- 0.83, -0.99 +- 1.36, 5.25 +- 0.71 cm,

all p b 0.05). HF patients also had significantly more leftwards and

lower, although not deeper, P_aorta than non-HF patients (1.63 +- 0.90,

1.16 +- 1.65, 5.43 +- 0.87 cm vs. 1.32 +- 0.67, 1.82 +- 1.43, 5.55 +-

0.80 cm). HF patients showed longer distance between P_max.LV and P_aorta (5.04 +- 0.69 cm) than non-HF patients (4.76 +- 0.64 cm, p = 0.013) (Table 1).

Subgroup analysis revealed only a few differences between HFpEF and HFrEF group. HFrEF group consisted of lesser females (48.6% vs. 74.4%, p = 0.038), lower EF (29.8 +- 6.5% vs. 55.7 +-

11.5%, p b 0.001), and shallower P_aorta z coordinate (5.23 +- 0.83

vs. 5.62 +- 0.86 cm, p = 0.046) than HFpEF group (Table 2).

Table 3 presents the sensitivity analysis to show the effect of the hand positioning height on the sternum. When the compressing hands were placed at the xiphisternal joint (y_sternum = 0 cm), the LV was com- pressed during CPR in 88.9% and 69.7% of the non-HF and HF patients, re- spectively. When the hands were located just 3 cm above the joint, the LV was compressed in only 18.1% and 14.5% of the non-HF and HF pa- tients, respectively. At 6 cm above the joint, the highest allowable posi- tion according to the current guidelines, all victims would have their aorta compressed directly during CPR rather than the LV.

Discussion

This study demonstrated that patients with HF had more leftward (6.9 mm), lower (5.2 mm), and deeper (5.1 mm) optimum compression site than those without HF. These seemingly marginal differences re- sulted in a considerable difference in the proportion of LV being com- pressed during CPR. If CPR is performed by compressing the midpoint of the lower half of the sternum based on the current guidelines, less HF patients (59.2%) would have their LV (vs. aorta) compressed than non-HF patients (77.8%). If the chest is compressed at just 3 cm above the xiphisternal joint, the aorta would be compressed, rather than the LV, in most of non-HF and HF patients (81.9% and 85.5%, respectively). At 6 cm above the joint, the upper allowable limit based on the current guidelines, no CA victims would have their LV compressed during CPR, regardless of the presence/absence of HF. (Table 3).

As far as we know, this study is the first to demonstrate that the op- timum compression site of HF patients differs from that of non-HF pa- tients. Furthermore, this study reaffirms the finding of the previous studies, which suggested that the ‘lowest possible’ part of the sternum should be compressed rather than its ‘lower half’ [8,10,17]. By conducting a sensitivity study using varying heights of the compressing hands on the sternum, we clearly revealed the critical effects caused by the seemingly small differences in hand positioning. Changing of the op- timum compression site to the ‘lowest possible’ part of the sternum is ap- plicable to any CA patients regardless of the presence/absence of HF, although HF patients may benefit more from this change than non-HF patients. The significance of the small difference in hand positioning

Fig. 5. Case enrolment. (CT, computed tomography; E, mitral peak velocity of early filling; e’, early diastolic mitral annulus velocity; EF, ejection fraction; LAVI, left atrium volume index; LVMI, left ventricle mass index; NT-proBNP, N-terminal pro-B type natriuretic peptide).

heights could be explained by focusing on the target of chest compres- sion. The ultimate target to compress during CPR is not the big thorax but the small LV beneath it. The lower half of the sternum may be too crude, a macroscopic term to define a compressing site that targets the smaller LV.

This study is also the first to introduce a 3D coordinate system on CT to delineate the accurate spatial location of the intrathoracic structures. This new frame enabled the performance of detailed analysis, calculation of geometric values, and widening of the compression site’s scope be- yond the sternum. This system may be used in other studies focusing on exact intrathoracic spatial relationships.

The results of this study are consistent with those of the previous studies. Shin et al., who controlled their study design strictly by instructing prospectively enrolled patients not to raise their arms during CT to simulate the CPR position, showed that the optimum compression point would be 2.5 +- 0.2 cm higher than the tip of the xiphoid process [8]. As the xiphoid process was 4.0 +- 0.2 cm long in this study, their op- timum compression point almost corresponds to the y_max.LV of our study, which was 1.0-1.5 cm lower than the xiphisternal joint (Table 1). Amelia et al. demonstrated that the aorta would be com- pressed rather than the LV in 38% of CA victims receiving CPR based on the current guidelines. Our results showed that the aorta was com- pressed in 22.2% and 40.8% of non-HF and HF cases, respectively [10]. Hwang et al. revealed that the stroke volume (SV) increases as the max- imally compressed point moves far away from the P_aorta towards the LV [9]. When this point was 3 cm away towards the LV, the SV was ap- proximately 50 cm³. When it was 3 cm towards the opposite direction within the aorta, it decreased to 20 cm³. Our study showed that the dis- tance of the P_max.LV from the P_aorta was 4.76 and 5.04 cm for non-HF and HF cases, respectively. Hence, the potential distance to maximize SV was increased from 3 to 5 cm.

This study has limitations. Firstly, some distances were immeasurable directly and calculated by multiplying the number of slices between the two points by the predetermined thickness of each slice (2, 3, or 5 mm). This approach adopted approximation in defining the edge of certain structures, implying potential error by half the thickness of each slice. Secondly, as the thorax is not cuboid, the planes on the 3D coordinate system do not match the real body surfaces. However, P_aorta and P_max.LV, the structures of interest, were located near the P_reference to lessen this mismatching problem.

Some authors argued that compressing anywhere on the sternum, the rigid bar hinging on the thoracic inlet, would result in the similar ef- fect [10]. However, as Shin et al. pointed out, sternal fracture, which abol- ishes this hypothesis, is a common complication of CPR [8,11,18]. Hwang et al. revealed that the distance of the maximally compressed point of the LV from the P_aorta significantly influenced on SV. This result also sup- ports that the compression site on the sternum matters [9].

Then, would it be practical to compress the fragile, narrow xiphoid process to maximize SV? There is one study which investigated the effect of compressing the lowest part of the sternum with 17 non-traumatic CA victims who had not gained spontaneous circulation within 30 min of CPR. It demonstrated that compressing at the lowest sternum led to higher peak arterial pressure during compression systole and end-tidal CO₂ pressure compared with the standard site [17]. However, no differ- ences were observed in arterial pressure during compression diastole and peak right atrial and coronary perfusion pressure between the two sites.

We suggest very carefully that we might need to go beyond the ster- num and compress the left lower anterior thorax just above the P_max.LV (P_max.LV’ in Fig. 4). As demonstrated, none of the P_sternum can surpass the P_max.LV’ in terms of closeness to the P_max.LV. There- fore, the most practical way to maximize SV during CPR, which is

Table 1

Demographic, clinical, laboratory, echocardiographic and CT characteristics according to the absence/presence of heart failure.

	Cases without HF	Cases with HF	p
	(n = 72)	(n = 76)
Demographics
Female, n (%)	40 (55.6)	47 (61.8)	0.54
Age (year)	55.9 +- 13.6	69.7 +- 13.3	b0.001??
Height (m)	1.61 +- 0.10 (n = 49)	1.58 +- 0.11 (n = 52)	0.14
Weight (kg)	63.7 +- 9.5 (n = 51)	60.0 +- 14.8 (n = 54)	0.13
Body mass index (kg/m2)	24.5 +- 3.5 (n = 49)	24.1 +- 4.6 (n = 51)	0.62
Comorbidity, n (%)
Hypertension	33 (45.8)	43 (56.6)	0.25
Diabetes mellitus	10 (13.9)	17 (22.4)	0.26
Hyperlipidemia	4 (5.6)	1 (1.3)	0.20
Stroke	4 (5.6)	12 (15.8)	0.08
Coronary artery disease	24 (33.3)	22 (28.9)	0.69
Chronic kidney disease	0 (0.0)	8 (10.5)	0.007??
Chronic liver disease	2 (2.8)	1 (1.3)	0.61
Chronic lung disease	5 (6.9)	19 (25.0)	0.006??
Tuberculosis	2 (2.8)	9 (11.8)	0.074
Malignancy	8 (11.1)	12 (15.8)	0.55
Main department, n (%)			0.005??
Cardiology	55 (76.4)	38 (50.0)
Pulmonology	5 (6.9)	19 (25.0)
Medicine other than the above	8 (11.1)	11 (14.5)
Surgery	4 (5.6)	8 (10.5)
NT-proBNP (pg/mL)	64 +- 46	6816 +- 10,862 (n = 73)	b0.001??
Echocardiography finding Ejection fraction (%)	66.6 +- 6.6	43.1 +- 16.1	b0.001??
E/e?	8.6 +- 2.3 (n = 66)	14.8 +- 6.8 (n = 48)	b0.001??
LA volume index (mL/m2)	24.3 +- 5.9 (n = 48)	44.1 +- 21.2 (n = 61)	b0.001??
LV mass index (g/m2)	95.9 +- 8.0 (n = 7)	144.2 +- 42.0 (n = 25)	b0.001??
CT finding
Length of the whole sternum (cm)	18.41 +- 2.31	18.11 +- 2.24	0.41
Length of the xiphoid process (cm)	4.20 +- 1.38	3.81 +- 1.32	0.078
LV (vs. aorta) beneath P_guideline, n (%)	56 (77.8)	45 (59.2)	0.025?
Coordinates of P_max.LV (cm)
x	5.00 +- 0.83	5.69 +- 0.98	b0.001??
y	-0.99 +- 1.36	-1.51 +- 1.67	0.041?
z	5.25 +- 0.71	5.76 +- 1.09	0.001??
Coordinates of P_aorta (cm)
x	1.32 +- 0.67	1.63 +- 0.90	0.016?
y	1.82 +- 1.43	1.16 +- 1.65	0.011?
z	5.55 +- 0.80	5.43 +- 0.87	0.40
Distance from P_max.LV to P_aorta (cm)	4.76 +- 0.64	5.04 +- 0.69	0.013?

CT, computed tomography; E, mitral peak velocity of early filling; e’, early diastolic mitral annulus velocity; HF, heart failure; LA, left atrium; LV, left ventricle; NT-proBNP, N-terminal pro-B type natriuretic peptide; P_aorta, the centre point of the aortic annulus, the ventriculo-aortic junction; P_guideline, the point at the lower one-fourth of the whole sternum, the midpoint of the lower half of the sternum; P_max.LV, the midpoint to bisect the maximal diameter of the left ventricle which is chosen by navigating sagittal images of the computed tomography.

* p b 0.05.

?? p b 0.01.

associated with better coronary perfusion pressure, earlier and success- ful achievement of return of spontaneous circulation (ROSC), higher mean aortic pressure at ROSC, and bigger EF after ROSC [9,19,20], would be to compress the P_max.LV’. By compression of the left lower anterior thorax, complications might occur, such as rib fracture, lung contusion, hemothorax, or pneumothorax, which might be even con- verted to Tension pneumothorax. However, based on the literature, even the current CPR practice, which recommends compression of the sternum, is accompanied by rib fracture in 13-97% of cases but with rare hemo- or pneumothorax (1.3-8.7%) [11,18]. The relatively rare de- velopment of hemo- or pneumothorax might be attributed to the soft- ness at the site of fracture. Fracture of the cartilaginous sternocostal joints caused by sternal compression according to the conventional CPR would leave relatively blunt surfaces causing less internal injury to the lung compared with fracture of the bony part of the ribs. P_max.LV’, which we cautiously suggest as an optimum compression site, is also located in these cartilaginous sternocostal joints and may re- sult in less internal injury similarly. If the benefits of maximizing the SV through compression of P_max.LV’ outweigh its complications or the complications could be minimized via new techniques, we may need to desert the sternum, the traditional default structure for chest

compression, and compress the P_max.LV’, which is located 5.0 cm left towards and 1.0 cm lower than the xiphisternal joint for non-HF patients and 5.7 cm left towards and 1.5 cm lower than the xiphisternal joint for HF patients.

Until further studies support these ideas, we may need to confine the compression site to the sternum. The result showed that the lower the compressing hands are located on the sternum, the more CA patients would have their LV compressed during CPR rather than the aorta (Table 1, Fig. 4). However, compression at or lower than the xiphisternal joint would inevitably cause dislocation and fracture to the feeble xi- phoid process, making chest compression less effective and increasing the risk of potentially fatal liver, stomach, or spleen injury. Therefore, the practical solution would be to compress the sternum as lowest pos- sible just above the xiphisternal joint.

Conclusion

The chest needs to be compressed as lowest as possible, such as just above the xiphisternal joint. This result is not confined to HF patients in CA, although it is especially applicable for them. Further study is need- ed to evaluate the risk and benefit of compressing the sternocostal joint

Table 2

Subgroup analysis within the heart failure patients regarding demographic, clinical, laboratory, echocardiographic and CT characteristics: HFpEF vs. HFrEF.

HFpEF (n = 39) HFrEF (n = 37) p

Demographics
Female, n (%)	29 (74.4)	18 (48.6)	0.038?
Age (year)	69.5 +- 13.0	69.9 +- 13.8	0.90
Height (m)	1.56 +- 0.09 (n = 28)	1.61 +- 0.12 (n = 24)	0.12
Weight (kg)	61.2 +- 16.8 (n = 30)	58.6 +- 12.0 (n = 24)	0.53
Body mass index (kg/m2)	25.2 +- 5.5 (n = 28)	22.9 +- 2.6 (n = 23)	0.055
Comorbidity, n (%)
Hypertension	24 (61.5)	19 (51.4)	0.51
Diabetes mellitus	8 (20.5)	9 (24.3)	0.90
Hyperlipidemia	1 (2.6)	0 (0.0)	1.00
Stroke	6 (15.4)	6 (16.2)	1.00
Coronary artery disease	13 (33.3)	9 (24.3)	0.54
Chronic kidney disease	3 (7.7)	5 (13.5)	0.48
Chronic liver disease	0 (0.0)	1 (2.7)	0.49
Chronic lung disease	10 (25.6)	9 (24.3)	1.00
Tuberculosis	3 (7.7)	6 (16.2)	0.30
Malignancy	4 (10.3)	8 (21.6)	0.30
Main department, n (%)			0.69
Cardiology	22 (56.4)	16 (43.2)
Pulmonology	8 (20.5)	11 (29.7)
Medicine other than the above	5 (12.8)	6 (16.2)
Surgery	4 (10.3)	4 (10.8)
NT-proBNP (pg/mL)	4603 +- 9298	9354 +- 12,059 (n = 34)	0.067
Echocardiography finding Ejection fraction (%)	55.7 +- 11.5	29.8 +- 6.5	b0.001??
E/e?	14.7 +- 6.3 (n = 30)	14.9 +- 7.7 (n = 18)	0.92
LA volume index (mL/m2)	42.1 +- 24.5 (n = 31)	46.2 +- 17.2 (n = 30)	0.45
LV mass index (g/m2)	134.2 +- 33.8 (n = 17)	165.3 +- 52.0 (n = 8)	0.085
CT finding
Length of the whole sternum (cm)	17.90 +- 1.93	18.32 +- 2.54	0.43
Length of the xiphoid process (cm)	3.86 +- 1.18	3.75 +- 1.46	0.71
LV (vs. aorta) beneath P_guideline, n (%)	22 (56.4)	23 (62.2)	0.78
Coordinates of P_max.LV (cm)
x	5.74 +- 1.02	5.64 +- 0.95	0.64
y	-1.68 +- 1.74	-1.32 +- 1.59	0.36
z	5.81 +- 1.26	5.71 +- 0.91	0.67
Coordinates of P_aorta (cm)
x	1.77 +- 0.90	1.49 +- 0.90	0.19
y	1.07 +- 1.65	1.27 +- 1.66	0.60
z	5.62 +- 0.86	5.23 +- 0.83	0.046?
Distance from P_max.LV to P_aorta (cm)	5.00 +- 0.69	5.09 +- 0.70	0.58

CT, computed tomography; E, mitral peak velocity of early filling; e’, early diastolic mitral annulus velocity; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; LA, left atrium; LV, left ventricle; NT-proBNP, N-terminal pro-B type natriuretic peptide; P_aorta, the centre point of the aortic annulus, the ventriculo-aortic junction; P_guideline, the point at the lower one-fourth of the whole sternum, the midpoint of the lower half of the sternum; P_max.LV, the midpoint to bisect the maximal diameter of the left ventricle which is chosen by navigating sagittal images of the computed tomography.

* p b 0.05.

?? p b 0.01.

area approximately 5-6 cm left towards and 1-1.5 cm lower than the xiphisternal joint to maximize the SV by compressing the maximum LV diameter.

Conflicts of interest

None to declare.

Table 3 Sensitivity analysis to calculate the effect of varying hand positioning heights on the sternum (P_sternum) on (1) the height difference between the sternum and P_aorta and (2) the pro- portion that the compression of the P_sternum is supposed to compress the LV rather than the aorta.

Height of P_sternum (cm) The height of P_aorta minus the height of P_sternum (cm) P_aorta is higher than P_sternum, n (%)

	Cases without HF (n = 72)	Cases with HF (n = 76)		Cases without HF (n = 72)	Cases with HF (n = 76)
6	-4.18 +- 1.43	-4.84 +- 1.65		0 (0)	0 (0)
5	-3.18 +- 1.43	-3.84 +- 1.65		2 (2.8)	1 (1.3)
4	-2.18 +- 1.43	-2.84 +- 1.65		4 (5.6)	2 (2.6)
3	-1.18 +- 1.43	-1.84 +- 1.65		13 (18.1)	11 (14.5)
2	-0.18 +- 1.43	-0.84 +- 1.65		30 (41.7)	23 (30.3)
1	0.82 +- 1.43	0.16 +- 1.65		48 (66.7)	37 (48.7)
0	1.82 +- 1.43	1.16 +- 1.65		64 (88.9)	53 (69.7)
-1	2.82 +- 1.43	2.16 +- 1.65		71 (98.7)	67 (88.2)
-2	3.82 +- 1.43	3.16 +- 1.65		71 (98.7)	75 (98.7)
-3	4.82 +- 1.43	4.16 +- 1.65		72 (100)	76 (100)

P_aorta, the centre point of the aortic annulus; P_sternum, the point where the compressing hand is in contact with the sternum.

Acknowledgement

The authors appreciate Dr. Won Sup Oh for his review of this manuscript.

References

1. Gorgels A. Out-of-hospital cardiac arrest-the relevance of heart failure. The Maas- tricht circulatory arrest Registry. Eur Heart J 2003;24:1204-9.
2. Chan MM, Lam CS. How do patients with heart failure with preserved ejection frac- tion die? Eur J Heart Fail 2013;15:604-13.
3. Cleland JG, Chattopadhyay S, Khand A, Houghton T, Kaye GC. Prevalence and inci- dence of arrhythmias and sudden death in heart failure. Heart Fail Rev 2002;7: 229-42.
4. Zile MR, Gaasch WH, Anand IS, Haass M, Little WC, Miller AB, et al. Mode of death in patients with heart failure and a preserved ejection fraction: results from the irbesartan in heart failure with preserved ejection fraction study (I-preserve) trial. Circulation 2010;121:1393-405.
5. Perkins GD, Handley AJ, Koster RW, Castren M, Smyth MA, Olasveengen T, et al. European resuscitation council guidelines for resuscitation 2015: section 2. Adult basic life support and Automated external defibrillation. Resuscitation 2015;95: 81-99.
6. Kleinman ME, Brennan EE, Goldberger ZD, Swor RA, Terry M, Bobrow BJ, et al. Part 5: adult basic life support and cardiopulmonary Resuscitation quality: 2015 American Heart Association guidelines update for cardiopulmonary resuscitation and emer- gency cardiovascular care. Circulation 2015;132:S414-35.
7. Japanese Circulation Society Resuscitation Science Study Group. Chest-compression- only bystander cardiopulmonary resuscitation in the 30:2 compression-to- ventilation ratio era. Nationwide observational study. Circ J 2013;77:2742-50.
8. Shin J, Rhee JE, Kim K. Is the inter-nipple line the correct hand position for effective chest compression in adult cardiopulmonary resuscitation? Resuscitation 2007;75: 305-10.
9. Hwang SO, Zhao PG, Choi HJ, Park KH, Cha KC, Park SM, et al. Compression of the left ventricular outflow tract during cardiopulmonary resuscitation. Acad Emerg Med 2009;16:928-33.
10. Pickard A, Darby M, Soar J. Radiological assessment of the adult chest: implications for chest compressions. Resuscitation 2006;71:387-90.
11. Buschmann CT, Tsokos M. Frequent and rare complications of resuscitation at- tempts. Intensive Care Med 2009;35:397-404.
12. Monsuez JJ, Charniot JC, Veilhan LA, Mougue F, Bellin MF, Boissonnas A. Subcapsular liver haematoma after cardiopulmonary resuscitation by untrained personnel. Re- suscitation 2007;73:314-7.
13. Berman M, Aravot D, Ben-Gal T, Sahar G, Sagie A, Vidne B. Cardiothoracic ratio: im- portant Prognostic tool in heart failure patients who are candidates for heart trans- plantation. Transplant Proc 2000;32:727-8.
14. Chon SB, Oh WS, Cho JH, Kim SS, Lee SJ. Calculation of the cardiothoracic ratio from portable anteroposterior chest radiography. J Korean Med Sci 2011;26:1446-53.
15. Ponikowski P, Voors AA, Anker SD, Bueno H, Cleland JG, Coats AJ, et al. 2016 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) developed with the special contribution of the Heart Failure Association (HFA) of the ESC. Eur Heart J 2016;37:2129-200.
16. Kelly JP, Mentz RJ, Mebazaa A, Voors AA, Butler J, Roessig L, et al. Patient selection in heart failure with preserved ejection fraction clinical trials. J Am Coll Cardiol 2015; 65:1668-82.
17. Cha KC, Kim HJ, Shin HJ, Kim H, Lee KH, Hwang SO. Hemodynamic effect of external chest compressions at the lower end of the sternum in cardiac arrest patients. J Emerg Med 2013;44:691-7.
18. Hoke RS, Chamberlain D. Skeletal chest injuries secondary to cardiopulmonary re- suscitation. Resuscitation 2004;63:327-38.
19. Klouche K, Weil MH, Sun S, Tang W, Povoas H, Bisera J. Stroke volumes generated by precordial compression during Cardiac resuscitation. Crit Care Med 2002;30: 2626-31.
20. Halperin HR, Lee K, Zviman M, Illindala U, Lardo A, Kolandaivelu A, et al. Outcomes from low versus high-flow cardiopulmonary resuscitation in a swine model of cardi- ac arrest. Am J Emerg Med 2010;28:195-202.