Comparative Evaluation of Three Transthoracic Echocardiographic Techniques for Cardiac Output and Stroke Volume Assessment Across Hemodynamic States

ABSTRACT

Objective

This study compared the accuracy of three transthoracic echocardiographic (TTE) techniques—fractional shortening (FS), left ventricular outflow tract velocity‐time integral (LVOT/VTI), and Simpson's method—for measuring cardiac output (CO) and stroke volume (SV) in hemodynamically unstable patients (systolic pressure < 90 mmHg), using pulse index continuous cardiac output (PiCCO) as the reference.

Methods

A retrospective analysis was conducted involving 12 patients admitted to an Emergency Intensive Care Unit between October 2023 and October 2024, who underwent a total of 54 echocardiographic examinations. The median length of hospital stay is 14 days. CO and SV values obtained using the three TTE methods as part of routine assessment were compared with simultaneous PiCCO measurements. The median number of examinations for patients is 4. Statistical analysis was performed using correlation methods.

Results

All three TTE methods demonstrated the capability to estimate CO and SV. The VTI method showed the highest overall accuracy (CO‐VTI vs. CO‐PiCCO: r = 0.950, p< 0.001). Under reduced CO conditions, correlations between echocardiographic and PiCCO‐derived measurements decreased for all methods; however, the VTI method maintained superior reliability (r = 0.606, p = 0.006). In contrast, the Simpson's method did not accurately reflect CO in this setting (SV‐Simpson vs. SV‐PiCCO: r = 0.408, p = 0.083). Notably, the performance of the VTI method remained consistent regardless of SV or heart rate variations (SV‐VTI vs. SV‐PiCCO: overall r = 0.970; with heart rate > 100 bpm, r = 0.946; with heart rate ≤ 100 bpm, r = 0.988).

Conclusion

The LVOT/VTI method exhibits the highest accuracy and consistency for measuring SV and CO, making it the preferred non‐invasive technique for hemodynamic evaluation in critically ill patients.

Comparison of three echocardiographic techniques for cardiac output and stroke volume assessment in critically ill patients. The LVOT/VTI method shows the highest accuracy and reliability compared with PiCCO.

Abbreviations

CO

Cardiac Output

FS

Fractional shortening

LVOT/VTI

Left ventricular outflow tract/velocity‐time integral

PAC

pulmonary artery catheter

PICCO

Pulse index Continuous Cardiac Output

SV

Stroke Volume

TTE

Transthoracic Echocardiography

1. Introduction

In adults at rest, cardiac output (CO) typically ranges from 4 to 8 L/min, while stroke volume (SV) falls between 60 and 100 mL/beat (Hall 2020). These parameters are essential for guiding hemodynamic management, particularly in the intensive care unit (ICU) setting (de Waal et al. ²⁰⁰⁹). Several modalities are currently employed to assess CO and SV, such as the pulmonary artery catheter (PAC), pulse index continuous cardiac output (PiCCO), and Doppler‐based ultrasound methods (Clement et al. 2017; Thiele et al. ²⁰¹⁵). Among these, PiCCO has been reported to provide superior accuracy in the measurement of CO and SV (Monnet and Teboul 2017; Sánchez‐Sánchez et al. ²⁰¹⁷).

Transthoracic echocardiography (TTE) offers a non‐invasive, portable, and safe means of evaluating cardiac function. It enables real‐time visualization, allows for repeated bedside measurements, and has demonstrated good agreement with PiCCO‐derived values for CO and SV in previous studies (Aslan et al. 2020; Gomes et al. ²⁰²⁴; Souto Moura et al. ²⁰¹⁸; Staer‐Jensen et al. ²⁰¹⁸; Grand et al. ²⁰²²). Frequently utilized TTE techniques for functional cardiac assessment include fractional shortening (FS) of the left ventricular short axis, the left ventricular outflow tract velocity–time integral (LVOT/VTI), and the Simpson method (Gunst et al. 2011).

Despite the established utility of TTE, limited evidence exists regarding the comparative accuracy of these three echocardiographic techniques in relation to PiCCO under conditions of variable CO. In particular, there is a lack of data on how changes in CO levels and heart rate affect the reliability of each method. The present study employed PiCCO as the reference standard to evaluate the accuracy of FS, LVOT/VTI, and Simpson's method in measuring CO and SV across a range of CO values, and to assess the influence of heart rate on measurement accuracy.

2. Participants and Methods

2.1. Study Participants

This retrospective study screened 242 patients admitted to the EICU between October 2023 and October 2024. Of these, 12 patients with hemodynamic instability (systolic pressure < 90 mmHg) underwent both PiCCO monitoring and multiple transthoracic echocardiographic (TTE) evaluations during their hospitalization. The cohort comprised six males and six females, with ages ranging from 60 to 93 years. Among the enrolled patients, seven were diagnosed with severe pneumonia, of whom four also presented with septic shock. Three patients were diagnosed with acute coronary syndrome; one had gastrointestinal bleeding complicated by heart failure; and one had multiple myeloma with associated multiple organ dysfunction syndrome. The average length of hospital stay was 15.9 days, with a median of 14 days. All 12 patients underwent internal jugular vein or subclavian vein catheterization and femoral artery thermodilution catheterization on the day of admission.

Exclusion criteria were as follows: inability to undergo TTE due to factors such as chest trauma, obesity, mechanical ventilation, or thoracic deformities; presence of conditions that could compromise the accuracy of TTE measurements, including atrial fibrillation, valvular heart disease, pulmonary embolism, or intracardiac shunts; periodic wall motion anomalies affecting the accuracy of FS method measurement results. Patients with severe peripheral arterial disease or arterial bypass grafts that could affect the precision of PiCCO‐derived data were also excluded.

2.2. Methods

A total of 54 TTE examinations were performed on the 12 patients enrolled in the study. Among them, three patients underwent three examinations, four underwent four, two underwent five, two underwent six, and one patient underwent seven examinations, yielding a median of four examinations per patient. Each examination involved the use of three echocardiographic methods to assess CO and SV. The echocardiographic measurements were subsequently compared with values obtained through pulse index continuous cardiac output (PiCCO) monitoring.

2.2.1. CO‐PiCCO and SV‐PiCCO Measurement Protocol

Cardiac output (CO‐PiCCO) and stroke volume (SV‐PiCCO) were measured using the PiCCO monitoring system (BeneVision N12, Mindray, China). The monitoring setup included insertion of a 7 Fr, 20 cm central venous catheter (CS‐27702‐E Arrowg + ard Blue Two‐Lumen CVC, Arrow International LLC, USA) into either the right internal jugular vein or the right subclavian vein. Concurrently, a 5 Fr, 20 cm thermodilution catheter equipped with a thermistor tip (PULSION Medical Systems SE, Germany) was inserted into the femoral artery.

During PiCCO monitoring, normothermia was maintained in all patients. The system was calibrated every 8 h using 20 mL of cold saline. Each calibration session included three consecutive thermodilution measurements, and the average of the three was recorded as the final value. CO‐PiCCO and SV‐PiCCO data were documented by a physician blinded to the echocardiographic results to ensure objective comparison (Gunst et al. 2011; Hewett ²⁰¹⁹; Wernly et al. ²⁰¹⁶; Gomes et al. ²⁰²⁴; Aslan et al. ²⁰²⁰).

2.2.2. TTE Measurement

TTE is a daily routine assessment item in the Intensive Care Unit of the author's Hospital. TTE assessments were performed by intensivists with formal training in critical care ultrasound, who had obtained relevant certification and demonstrated proficiency in multiple techniques for evaluating cardiac function. All examinations were conducted using a TE7 ultrasound system (Mindray, Shenzhen, China) equipped with a P4‐2s cardiac probe. Imaging parameters, including imaging depth and gain, were adjusted as necessary to optimize image quality. For each participant, sequential assessments were conducted using three echocardiographic methods—fractional shortening (FS), left ventricular outflow tract velocity‐time integral (LVOT/VTI), and Simpson's method—yielding the following parameters: CO‐FS, CO‐VTI, CO‐Simpson, SV‐FS, SV‐VTI, and SV‐Simpson (Gunst et al. 2011).

2.2.2.1. Fractional Shortening (FS) Method

Using the parasternal long‐axis view, an M‐mode cursor was positioned at the distal end of the mitral valve to acquire a one‐dimensional cross‐sectional view of the left ventricle. Left ventricular end‐systolic and end‐diastolic diameters were measured, and stroke volume (SV‐FS) was automatically calculated by the ultrasound system. Cardiac output (CO‐FS) was subsequently determined by multiplying SV‐FS by the heart rate.

2.2.2.2. LVOT/VTI Method

The left ventricular outflow tract (LVOT) diameter was measured in the parasternal long‐axis view. Pulsed‐wave Doppler was then applied in the apical five‐chamber view to obtain the velocity–time integral (LVOT‐VTI) across the aortic valve. Stroke volume (SV‐VTI) was calculated using the formula: SV‐VTI = LVOT area × LVOT‐VTI, where LVOT area = π × (LVOT radius)². CO‐VTI was calculated by multiplying SV‐VTI by the heart rate (CO‐VTI = heart rate × LVOT‐VTI × LVOT radius² × 3.14). For participants in sinus rhythm, values were averaged over 2–3 cardiac cycles. In the presence of atrial fibrillation, measurements were averaged over 5–7 cardiac cycles to account for beat‐to‐beat variability (Orde et al. 2017).

2.2.2.3. Simpson's Method

In the apical four‐chamber view, the end‐systolic and end‐diastolic contours of the left ventricle were manually traced. The ultrasound system reconstructed the left ventricle as a series of stacked cylindrical segments of equal height but varying diameters. SV‐Simpson was calculated as the difference between the estimated end‐systolic and end‐diastolic volumes. CO‐Simpson was then derived by multiplying SV‐Simpson by the heart rate.

2.2.3. Statistical Methods

All data collected in this study were analyzed using R software (Version 4.4.1, R Foundation for Statistical Computing, Vienna, Austria). Linear regression with restricted cubic splines (RCS) was used to determine optimal cutoff values. Pearson's linear correlation analysis was applied to assess linear relationships among continuous variables, with the strength of correlation interpreted as follows: coefficients with absolute values ranging from 0.7 to 1.0 were considered strong. The significance threshold was set at α = 0.05. In cases where the assumption of normality was not satisfied, Spearman's rank correlation analysis was used. A p value < 0.05 was considered statistically significant.

3. Results

3.1. Relationship Between CO‐PiCCO and CO‐FS, CO‐VTI, and CO‐Simpson

Cardiac output measurements obtained via PiCCO (CO‐PiCCO) demonstrated statistically significant positive correlations with all three TTE methods: CO‐FS, CO‐VTI, and CO‐Simpson. All correlations were statistically significant, with p values < 0.001. The correlation between CO‐PiCCO and CO‐VTI demonstrated the highest strength (Table 1). A greater absolute value of the correlation coefficient reflected a stronger association.

TABLE 1.

Spearman's correlation analysis between CO‐PiCCO and CO‐FS, CO‐VTI, and CO‐Simpson.

	r/p	CO‐FS	CO‐VTI	CO‐simpson
CO‐PICCO	r	0.882	0.950	0.856
	p	0.000	0.000	0.000
CO‐PICCO ≥ 4	r	0.755	0.940	0.637
	p	0.000	0.000	0.000
CO‐PICCO < 4	r	0.583	0.606	0.408
	p	0.009	0.006	0.083

Note: p < 0.05 indicates statistical significance. The Spearman's correlation coefficient (r) represents the strength and direction of the relationship. Positive values indicate a direct correlation and negative values indicate an inverse relationship. Higher absolute values of r signify stronger correlations.

Using the lower limit of normal cardiac output (4 L/min) as a cutoff point (Hall 2020), a statistically significant non‐linear relationship was identified between CO‐VTI and CO‐PiCCO (p < 0.001) (Figure 1). When CO‐PiCCO values were ≥ 4 L/min, the correlation coefficient between CO‐PiCCO and CO‐VTI reached 0.940, indicating a strong correlation. In cases where CO‐PiCCO was < 4 L/min, the correlation coefficients between CO‐PiCCO and both CO‐FS and CO‐VTI decreased but remained statistically significant. In contrast, the correlation between CO‐PiCCO and CO‐Simpson was not statistically significant (p = 0.083; Table 1).

FIGURE 1.

Linear correlation between CO‐VTI and CO‐PiCCO.

Overall, all three TTE methods demonstrated the capability to assess CO, with the VTI method exhibiting the highest accuracy. Under conditions of reduced cardiac output, correlation strengths declined for all methods; however, the VTI method maintained superior accuracy, whereas the Simpson method did not accurately reflect CO.

3.2. Relationship Between SV‐PiCCO and SV‐FS, SV‐VTI, and SV‐Simpson

Stroke volume values obtained through PiCCO monitoring (SV‐PiCCO) demonstrated positive correlations with all three echocardiographic methods: SV‐FS, SV‐VTI, and SV‐Simpson. All correlations reached statistical significance. Among the three echocardiographic methods, the strongest correlation was observed between SV‐PiCCO and SV‐VTI, with a correlation coefficient of 0.970 (Table 2).

TABLE 2.

Pearson's correlation analysis between SV‐PiCCO, SV‐FS, SV‐VTI, SV‐Simpson, and heart rate.

	r/p	SV‐FS	SV‐VTI	SV‐Simpson
SV‐PICCO	r	0.936	0.970	0.937
	p	0.000	0.000	0.000
SV‐PICCO ≥ 60	r	0.560	0.830	0.767
	p	0.000	0.000	0.000
SV‐PICCO < 60	r	0.904	0.944	0.878
	p	0.000	0.000	0.000

Note: p < 0.05 indicates statistical significance. The Pearson's correlation coefficient (r) denotes the strength and direction of the relationship. Positive r values reflect a positive correlation, and negative values indicate a negative correlation. Higher absolute values of r signify stronger correlations.

Using 60 mL/beat—the lower limit of the normal range for stroke volume—as a threshold, a statistically significant non‐linear relationship was observed between SV‐VTI and SV‐PiCCO (p < 0.001) (Figure 2) (Hall 2020). The SV‐VTI method consistently demonstrated the strongest correlation with SV‐PiCCO, regardless of whether SV‐PiCCO values were above or below the 60 mL/beat threshold (Table 2).

FIGURE 2.

Linear correlation between SV‐VTI and SV‐PiCCO.

3.3. Relationship Between SV‐PiCCO and SV‐FS, SV‐VTI, and SV‐Simpson at Different Heart Rates

Using 100 beats per minute—the upper limit of normal resting heart rate—as a threshold, patients were categorized into two groups: a tachycardia group (heart rate > 100 bpm) and a normal heart rate group (heart rate ≤ 100 bpm) (Hall 2020). In both groups, the correlation coefficient between SV‐PiCCO and SV‐VTI exceeded the correlations observed with the SV‐FS and SV‐Simpson methods (Table 3).

TABLE 3.

Pearson's correlation analysis of SV‐PiCCO, SV‐FS, SV‐VTI, SV‐Simpson, and heart rate, stratified by heart rate > 100 and ≤ 100 bpm.

		r/p	SV_FS	SV_VTI	SV_Simpson
Heart rate > 100 bpm	SV_PICCO	r	0.917	0.946	0.831
		p	0.000	0.000	0.000
Heart rate ≤ 100 bpm	SV_PICCO	r	0.880	0.988	0.967
		p	0.000	0.000	0.000

Note: p < 0.05 indicates statistical significance. The Pearson's correlation coefficient (r) denotes the strength and direction of the relationship. Positive values reflect a positive correlation and negative values indicate a negative correlation. Higher absolute values of r signify stronger correlations.

4. Discussion

In the intensive care unit (ICU) setting, reduced CO and hemodynamic instability—whether resulting from cardiac pathology or other underlying conditions—represent critical determinants of patient mortality (Huang 2019). Although the pulmonary artery catheter (PAC) remains the gold standard for comprehensive hemodynamic assessment, its clinical use has declined due to the associated risks, including thromboembolism, infection, and arrhythmias, along with evidence linking PAC use to increased mortality (Horster et al. 2012; Meersch et al. ²⁰¹⁶; Sandham ²⁰⁰⁴; Richard et al. ²⁰¹¹; Lee et al. ²⁰¹⁷). The development and widespread adoption of minimally invasive methods for CO measurement have further contributed to the reduced use of PAC contemporary clinical practice (Gomes et al. 2024).

The PiCCO system was the first pulse contour device implemented in clinical settings for CO measurement (Messina et al. 2023). Multiple studies have demonstrated a strong correlation between PiCCO‐derived values and those obtained through PAC (Monnet and Teboul 2017; Litton and Morgan ²⁰¹²). The PiCCO system calculates CO by analyzing the thermodilution curve recorded at the arterial catheter tip following the administration of a cold saline bolus into the central venous circulation. The intermittent thermodilution data are then used to calibrate the pulse contour algorithm, enabling continuous and real‐time monitoring of CO—an advantage not offered by PAC (Monnet and Teboul 2017; Beurton et al. ²⁰¹⁹).

Despite the clinical utility of pulse wave contour devices, their reliance on invasive procedures, potential for complications, and high associated costs have limited their widespread adoption. In recent years, focused echocardiography has gained prominence as a preferred modality for hemodynamic monitoring due to its non‐invasive nature. This approach eliminates the need for arterial and central venous catheterization, thereby reducing the risk of complications associated with invasive techniques (Monnet and Teboul 2017; Messina et al. ²⁰²³).

TTE offers several advantages, including non‐invasiveness, portability, safety, and the ability to provide real‐time results through repeated bedside measurements. In our center, ultrasound is widely used as a routine examination method for patients' daily examinations without increasing patient costs. Previous studies have demonstrated a strong correlation between TTE‐derived and PiCCO‐derived measurements of CO and SV (Aslan et al. 2020). TTE is particularly effective in detecting abnormalities in ventricular function, valvular pathology, and chamber dimensions (Price et al. 2008). However, concerns have been raised regarding its accuracy in quantifying CO when compared to invasive methods (Soliman‐Aboumarie et al. ²⁰²²; Jensen et al. ²⁰⁰⁴). In the ICU setting, the accuracy of results may be affected by various factors such as mechanical ventilation, the presence of drainage tubes, wound dressings, and suboptimal patient positioning, which can influence image quality (Orde et al. 2017; Fletcher and Grounds ²⁰¹²). Furthermore, each of the three TTE‐based methods used for estimating SV and CO possesses distinct advantages and limitations, necessitating consideration of clinical context and technical constraints when interpreting findings.

Left ventricular ejection fraction, as assessed by the FS method, is commonly used as an indicator of systolic function and has been widely regarded as a conventional reference in the field of cardiology, with its clinical utility supported by multiple studies (Orde et al. 2017; Lang ²⁰¹⁶; Silbiger and Singh ²⁰¹⁶). The findings of the present study demonstrated a strong correlation between CO‐FS and CO‐PiCCO, as well as between SV‐FS and SV‐PiCCO, which aligns with previous literature. However, reduced SV and CO inherently reflect impaired ejection fraction (EF), as these parameters may be influenced by factors such as increased left ventricular end‐diastolic volume, segmental wall motion abnormalities, and tachycardia (Gardner et al. 2009).

The FS method exhibited reduced accuracy under conditions of low CO, whereas its performance improved in the presence of low SV, a pattern consistent with the observations reported by Gardner et al. (2009) Notably, the accuracy of FS‐derived measurements appeared to improve during tachycardia, an outcome that contrasts with the findings of Gardner et al. This apparent discrepancy may be explained by the methodological characteristics of the FS technique, which involves M‐mode measurements at the mid‐ventricular level. In the presence of segmental wall motion abnormalities, the relative contributions to cardiac ejection from the mid‐ventricular and apical regions may differ. Tachycardia, by reducing SV, may diminish these regional disparities, potentially enhancing the accuracy of FS‐derived measurements under such conditions.

The Simpson method calculates total ventricular volume by dividing the left ventricle into multiple thin slices, avoiding reliance on geometric assumptions. This characteristic makes the method less susceptible to the influence of cardiac motion and particularly applicable in cases involving morphologically abnormal hearts (Voigt et al. 2015). The Simpson method can be extended by relying on operator experience, good image quality for optimal visualization of endocardial boundaries, geometric assumptions, load dependence, and poor reliability among raters (Marwick 2018; Otterstad et al. ¹⁹⁹⁷; Sonaglioni et al. ²⁰²⁴). In theory, this technique offers a more accurate assessment of left ventricular ejection function, a finding supported by the present study. However, its accuracy is highly dependent on precise delineation of the endocardial border, limiting its application to experienced operators and to non‐emergent clinical settings (Lang et al. 2005; Akil et al. ²⁰²⁵). The relatively low correlation between Simpson‐derived and PiCCO‐derived measurements observed in this study supports these limitations.

Previous research has indicated that, under conditions of low cardiac output, impaired left ventricular function, chamber dilation, or compromised tissue perfusion may contribute to reduced ultrasound image quality, thereby increasing the likelihood of measurement error when using the Simpson method (Thavendiranathan et al. 2013). The current study similarly demonstrated a diminished correlation between Simpson‐derived values and reference measurements in low CO conditions. Furthermore, the accuracy of this method relies on acquiring optimal apical four‐chamber views at precise points during the cardiac cycle—namely end‐diastole and end‐systole (Fleisher et al. 2006). In the presence of tachycardia, the abbreviated interval between systole and diastole may lead to greater timing discrepancies during image acquisition, thereby reducing measurement accuracy (Nagueh et al. 2016). This limitation was corroborated in the present study, as reduced accuracy of SV‐Simpson values was observed among patients with elevated heart rates.

The LVOT‐VTI method has been recognized as a reliable approach for measuring cardiac function in the majority of critically ill patients. It has demonstrated superior reproducibility in evaluating left ventricular systolic function among patients undergoing mechanical ventilation and those experiencing hemodynamic instability (Dinh et al. 2012). Multiple studies have reported the advantages of the LVOT‐VTI method in cardiac function assessment (Blanco 2020; Tan et al. ²⁰¹⁷). In the present study, the LVOT‐VTI method outperformed the FS and Simpson methods.

In critically ill populations, reductions in CO are often multifactorial, commonly resulting from a combination of cardiac dysfunction and tachycardia. Under such conditions, structural distortion of the heart or the presence of aortic valve regurgitation may introduce measurement inaccuracies. These inaccuracies may be further amplified due to the inherent characteristics of the VTI calculation method (Lang et al. 2015). Consequently, bias in LVOT‐VTI results may occur. It is important to note that tachycardia does not necessarily reflect reduced CO. A study by Gunst et al. (2011) demonstrated that the LVOT‐VTI method is minimally influenced by heart rate and maintains accuracy in evaluating cardiac ejection function even at elevated heart rates, consistent with the findings observed in the present study.

There were also several limitations in this manuscript. First, this is a retrospective study with a small sample size. Second, despite the advantages of LVOT‐VTI, its accuracy is contingent on the proper acquisition of the apical five‐chamber view. Cardiac structural changes and variations in the aortic ejection angle may cause misalignment between the Doppler sampling line and the true direction of blood flow, thereby resulting in measurement deviations. These deviations tend to become more pronounced at higher CO levels, which may account for the slightly reduced correlation observed in patients with high SV in the present analysis. Nevertheless, overall cardiac function exerted minimal impact on the reliability of LVOT‐VTI measurements, a finding consistent with observations reported by Gunst et al. (Gunst et al. 2011). From a clinical perspective, SV may serve as a more accurate indicator of cardiac function than CO.

5. Conclusion

In summary, TTE demonstrated good agreement with invasive monitoring in the assessment of cardiac function. Among the three echocardiographic assessment methods evaluated, the VTI method demonstrated superior performance. Across varying levels of SV and in the presence of tachycardia, the VTI method consistently provided accurate evaluations of cardiac function. These findings support the use of the VTI method as the preferred non‐invasive approach for SV evaluation in critically ill patients.

6. Limitations

This study was limited to patients with hemodynamic instability, and variations in baseline cardiac function may have influenced the observed outcomes. The enrollment was confined to a 1‐year period, which limited the number of patients eligible for PiCCO monitoring. Furthermore, the resulting relatively small sample size restricts the generalizability of the findings. Future investigations with larger sample sizes and stratified subgroup analyses are warranted to validate these findings. Additionally, the potential impact of vasoactive medications, particularly inotropic agents, was not addressed and should be considered in subsequent research. The use of a prospective study design would further strengthen the reliability and clinical applicability of the results and is recommended for future studies.

Author Contributions

Conception and design of the research: Lei Wang, Da‐Wei Wang, Rui‐Ming Xu, Jing Zhang. Acquisition of data: Lei Wang, Rui‐Ming Xu, Jian‐Zhong Li. Analysis and interpretation of the data: Lei Wang, Rui‐Ming Xu, Jian‐Zhong Li. Statistical analysis: Lei Wang, Da‐Wei Wang. Writing of the manuscript: Lei Wang. Critical revision of the manuscript for intellectual content: Da‐Wei Wang, Jing Zhang. All authors read and approved the final draft.

Funding

The authors have nothing to report.

Ethics Statement

This study was approved by the Institutional Review Board of Beijing Tongren Hospital, Capital Medical University. The requirement for informed consent was waived by the IRB due to the retrospective nature of the study and the use of de‐identified data.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

All data generated or analysed during this study are included in this article. Further enquiries can be directed to the corresponding author.